Received: 21 December 2023

DOI: 10.1049/qtc2.12094

REVIEW
- -Revised: 21 February 2024 Accepted: 23 March 2024

- IET Quantum Communication

Advances in artificial intelligence and machine learning for

quantum communication applications

Mhlambululi Mafu

Department of Physics, Case Western Reserve Abstract

University, Cleveland, Ohio, USA
Artificial intelligence (AI) and classical machine learning (ML) techniques have revolu-
tionised numerous fields, including quantum communication. Quantum communication
Correspondence
Mhlambululi Mafu.
technologies rely heavily on quantum resources, which can be challenging to produce,
Email: mxm1625@[Link] control, and maintain effectively to ensure optimum performance. ML has recently been
applied to quantum communication and networks to mitigate noise‐induced errors and
analyse quantum protocols. The authors systematically review state‐of‐the‐art ML ap-
plications to advance theoretical and experimental central quantum communication
protocols, specifically quantum key distribution, quantum teleportation, quantum secret
sharing, and quantum networks. Specifically, the authors survey the progress on how ML
and, more broadly, AI techniques have been applied to optimise various components of a
quantum communication system. This has resulted in ultra‐secure quantum communi-
cation protocols with optimised key generation rates as well as efficient and robust
quantum networks. Integrating AI and ML techniques opens intriguing prospects for
securing and facilitating efficient and reliable large‐scale communication between multiple
parties. Most significantly, large‐scale communication networks have the potential to
gradually develop the maturity of a future quantum internet.

KEYWORDS
learning (artificial intelligence), quantum communication, quantum cryptography, quantum information

1 | INTRODUCTION 9]. Owing to these properties, any attempt by an eavesdropper

to access information encoded in quantum states can be
Quantum communication is the art of transferring a quantum detected before confidential information is exchanged between
state from one place to another [1, 2]. Quantum states encode the legitimate parties. Classical communication relies on bits to
quantum information and allow one to perform tasks that encode and transmit data, and its security is guaranteed by the
afford secure and efficient communication surpassing the ca- computational hardness of specific mathematical problems,
pabilities of classical information [3, 4]. Quantum communi- such as factorisation of large integers [10, 11]. The advances in
cation harnesses the distinct characteristics of quantum quantum computing pose a significant threat to classical al-
mechanics, such as the phenomenon of quantum entanglement gorithms based on symmetric and asymmetric cryptography,
[5], when two objects can be instantaneously connected; su- which are only computationally secure [12]. To break public‐
perposition [6], a case in which a single object can be in two key cryptography, one needs the ability to solve problems
places at once; the uncertainty principle [7], which states that that are believed to be hard for classical and quantum com-
there is inherent uncertainty in measuring a variable of a par- puters. For instance, Shor's algorithm, a quantum computing
ticle, and the quantum no‐cloning theorem [8], which prevents algorithm can efficiently factor large numbers, breaking the
perfect copying of an arbitrary unknown quantum state to security of the Rivest‐Shamir‐Adleman (RSA) encryption al-
enable the secure and efficient transmission of information [3, gorithm [13]. The RSA algorithm is widely used for secure data

-
This is an open access article under the terms of the Creative Commons Attribution License, which permits use, distribution and reproduction in any medium, provided the original work is
properly cited.
© 2024 The Authors. IET Quantum Communication published by John Wiley & Sons Ltd on behalf of The Institution of Engineering and Technology.

202 IET Quant. Comm. 2024;5:202–231. [Link]/journal/qtc2

 26328925, 2024, 3, Downloaded from [Link] by Pondicherry University, Wiley Online Library on [10/08/2025]. See the Terms and Conditions ([Link] on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
MAFU
- 203

transmission and digital signatures, and its susceptibility to Recently, ML and more broadly, AI techniques have been
Shor's algorithm potentially presents a critical challenge to successfully applied to various aspects of quantum communi-
cybersecurity. cations, such as identifying fundamental quantum protocols
Quantum communication promises information‐theoretic [55], including quantum teleportation [62, 63], entanglement
or unconditional security because it relies on the laws of purification [64–67], quantum repeaters [68, 69] and estimating
physics rather than the presumed computational hardness of parameters of quantum devices [70–72]. AI and ML techniques
certain problems [3, 9]. The secure exchange of sensitive in- have demonstrated outstanding performance in addressing
formation could be transmitted via a quantum network, such as noise, channel losses, and quantum state detection [55, 57].
quantum photonic chips [14, 15], optical fibre, or free‐space Since quantum communication is highly susceptible to errors
channels [16, 17]. One of the main goals of quantum due to noise and decoherence caused by environmental factors
communication is connecting quantum computers to build a and hardware imperfections [4, 73], AI and ML promises to
quantum network [18]. This could increase the total computing significantly contribute to addressing classical and quantum
power, especially if only processors with a few qubits are error correction [74–78]. Specifically, ML algorithms allow the
available at each network node [19] and will enable the real- optimisation of quantum systems and possibilities to identify
isation of the envisaged quantum internet [20–22]. Neverthe- and correct errors more effectively [79–81]. This will improve
less, the limitations faced by quantum information carriers, the reliability and accuracy of quantum communication sys-
such as losses at large distances or quantum decoherence, a tems [42, 75, 82–85].
quantum network will also contain other elements, such as a The field of AI and ML dates back to the 1950s when Alan
quantum repeater or quantum switch to enable the trans- Turing proposed developing intelligent machines and testing
mission of quantum information over long distances without their intelligence [86] and at the Dartmouth Summer Research
degradation [21, 23]. Most significantly, the primary paradigms Project on Artificial Intelligence, a summer workshop widely
of quantum communication are quantum key distribution credited as the founding event of AI as a field [87, 88]. On the
(QKD) [16], quantum teleportation [24], quantum secret other hand, quantum communication has existed since the 1980s
sharing (QSS) [25] and quantum networks [20]. Notably, when Bennett and Brassard proposed a quantum physics‐based
quantum communication technologies such as QKD for point‐ solution to the key distribution problem [9, 16]. While these
to‐point connection at short distances or by using trusted re- fields have long existed, the application, development of AI and
peaters at large distances have gone beyond prototypes and ML, and impact on quantum communication technologies are
begun being commercialised [26–28]. Furthermore, the Inter- still relatively low, though it is proliferating [82, 89]. Conse-
national Organization for Standardisation has developed quently, we present a concise overview of recent developments
standards governing the security requirements, test, and eval- regarding the main paradigms of quantum communication,
uation methods for QKD [29, 30]. This versatility of quantum including quantum networks, as well as a brief discussion of
communication demonstrates its immense potential for some of the main challenges. Specifically, we discuss how AI and
transforming the modern world, ranging from enhancing se- ML algorithms have been integrated into various components of
curity to revolutionising long‐distance communication to quantum communication systems in order to overcome imple-
paving the way for a quantum internet [21]. The quantum mentation challenges. We advise the reader that due to the
internet is expected to profoundly impact various technological subjective nature of the selection process and the pace at which
frontiers and everyday life through providing quantum secure AI and ML applications and developments progress, some
cloud‐based services [22, 31, 32]. Clearly, in addition to its contributions will inevitably be missed. We do not conduct an
technological significance, quantum communication protocols extensive review of various quantum communication protocols.
and networks are expected to have far‐reaching social, eco- However, we provide a comprehensive list of literature on sig-
nomic, political, and geostrategic implications [33, 34]. nificant contributions to the field, covering the more detailed
Meanwhile, advances in artificial intelligence (AI) and issues that we can only briefly cover in a review. After discussing
machine learning (ML) have revolutionised science and society the examples in our article, we hope the reader will appreciate
[35]. Precisely, AI and ML enable machines to perform com- and understand the significance of AI and ML techniques in
plex tasks autonomously, without human intervention [36, 37]. quantum communication protocols and quantum networks.
For instance, AI and ML promise unprecedented applications
in data processing problems such as data prediction, data
sorting, pattern recognition, and classification [36, 38–41]. 2 | FUNDAMENTALS OF AI AND ML
These technologies have already begun transforming various
sectors such as banking and finance [42, 43], business [44, 45], AI refers to the development of computer systems with the
consumer [46, 47], education [48–51] and defence [52, 53]. capacity to solve advanced problems based on analytical
Furthermore, AI and ML applications have become essential in models that generate predictions, rules, answers, recommen-
communication, resulting in improved, efficient, and secure dations, or similar outcomes [37, 90–92]. These systems are
systems [54–60]. A suitable example is Rivest's 1991 discussion designed to analyse data, recognise patterns, and make de-
on the relationship between cryptography and ML, high- cisions or predictions based on the information they gather
lighting how both fields have developed ideas and techniques [93, 94]. ML is a subset of AI that focuses on developing al-
from each other resulting in improved security systems [61]. gorithms and models that enable computers to learn from and
 26328925, 2024, 3, Downloaded from [Link] by Pondicherry University, Wiley Online Library on [10/08/2025]. See the Terms and Conditions ([Link] on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
204
- MAFU

make predictions or decisions without being explicitly pro- 3 | APPLICATIONS OF AI AND ML TO

grammed [94, 95]. The algorithms also adapt in response to QUANTUM COMMUNICATION
new data and experiences to improve their efficacy over time
[96]. With the capacity of systems to process vast amounts of On the one hand, AI and ML have become ubiquitous and
data and make intelligent decisions, AI and ML have opened effective techniques for data processing and classification, and
up new unprecedented possibilities and opportunities in on the other hand, the ability to manipulate and control
various fields [97, 98], including physical sciences [40, 99, 100] quantum states has resulted in a wide range of applications that
and more specifically, quantum communications. The ML al- enable the secure and efficient transmission of information
gorithms can be grouped into supervised and unsupervised through quantum systems. The increasing complexity of
learning [101]. The adjectives ‘supervised or unsupervised’ quantum systems and communication technologies necessi-
indicate whether there are labelled samples in the database [93, tates powerful data processing and analysis tools [119]. One of
96]. Moreover, reinforcement learning has emerged as a new the most important developments is integrating or co‐existing
ML category inspired by behavioural psychology and neuro- quantum signal channels into classical telecommunication
science [102]. Reinforcement learning concerns an agent's networks. While significant progress has been reported, various
specific form of reward or utility connected to its environment implementation challenges remain [120]. Most significantly, AI
via perception and action [103]. A subset of ML called deep and ML have proven valuable tools for addressing these
learning (DL), allows a computational model composed of challenges in recent years. The reason for this is that AI and
multiple layers of processing units to learn multiple levels of ML techniques thrive at learning highly complex input–output
abstraction in given data [93, 104]. Most significantly, DL is mappings, which enables optimisation of systems [121, 122],
excellent at discovering intricate structures in high‐dimensional developing signalling and detection schemes for complex
data and is therefore applicable to many domains of science channels or channels without analytical or numerical models
[105] and business [106]. The relationship among the AI, ML, [123], and detecting ultra‐sensitive signals [124, 125]. Due to
and DL techniques is shown in Figure 1. the unprecedented progress in these fields, in the following
The family of AI and ML algorithms can be categorised subsections, we examine how AI and ML techniques have been
based on their similarity in functionality and structure. This has applied to solve various theoretical and practical challenges in
given rise to regression algorithms [107], instance‐based algo- QKD, quantum teleportation, QSS, and quantum networks. To
rithms [108], regularisation algorithms [109], decision tree (DT) provide context, we offer a concise overview of the various
algorithms [110], Bayesian algorithms [111], clustering algo- quantum communication paradigms. This is followed by
rithms [112, 113], artificial neural networks (ANNs) [114, 115], examining the current challenges and briefly discussing how
DL algorithms [98], long short‐term memory (LSTM) algo- recent advances in AI and ML have been used to address them,
rithms [116], dimension reduction algorithms [117], and thus enhancing quantum communication applications.
ensemble algorithms [118]. We discuss how the synergistic
integration of these AI and ML techniques has enhanced
quantum communication system capabilities, resulting in more 3.1 | Quantum key distribution
secure, reliable, and efficient quantum information transmission.
QKD‐the establishment of information‐theoretically secure
random key, which guarantees a level of security that is inde-
pendent of any computational assumptions based on quantum
mechanics [3, 11, 16, 126]. The information about a secure key
is transmitted using qubits. The symmetric cryptographic sys-
tems such as the one‐time pad utilise the securely generated
secret keys for encryption and decryption of messages to
ensure confidentiality and prevent eavesdropping [9, 127, 128].
The secret key enables secure communication and authenti-
cation between spatially separated legitimate parties, known as
Alice and Bob. QKD protocols require a quantum channel,
through which qubits containing information about the
distributed key are exchanged, and a public channel, which is
used to check whether communication through the quantum
channel is distorted. An illustration of a general QKD protocol
where Alice and Bob share an insecure quantum channel and
F I G U R E 1 An overview of the relationship between AI, ML, and DL. an authenticated classical channel wishes to share a secret
A classification algorithm and a regression algorithm are examples of secure key in the presence of an adversarial eavesdropper, Eve
supervised learning, while a clustering algorithm is an example of
is shown in Figure 2. While Eve can intercept a quantum
unsupervised learning. Notably, DL is a subset of ML in which multilayered
neural networks learn from data and examples include convolutional neural channel and extract information by measuring the transmitted
networks and recurrent neural networks. AI, artificial intelligence; DL, deep quantum states, she is prevented from doing so by the laws of
learning; ML, machine learning. quantum mechanics, since any measurement, in general,
 26328925, 2024, 3, Downloaded from [Link] by Pondicherry University, Wiley Online Library on [10/08/2025]. See the Terms and Conditions ([Link] on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
MAFU
- 205

QKD has two significant characteristics: practical security

and key‐rate performance [140, 141]. These characteristics
enable or limit QKD use in real‐life applications [17, 132, 133,
142]. The key rates of DV‐QKD protocols are evaluated using
analytical methods since most of them are symmetric [9, 143–
147]. However, symmetry assumptions usually fail for other
classes of protocols because of the differences in encoding.
This leads to new challenges in analysing the security of such
protocols. As a result, numerical methods based on convex
F I G U R E 2 The schematic setting of a typical QKD protocol. The two optimisations have been developed [148–151]. Typically, a
legitimate parties, Alice and Bob are connected by a quantum channel, QKD system consists of three components: a source, a
which an eavesdropper, Eve can tap without any restriction other than by
channel, and a measurement device [3]. The source is relatively
quantum mechanics. In the authenticated public classical channel, Eve can
only listen to the conversation but cannot modify the messages being sent simple and can be well characterised [152, 153]. The security of
between Alice and Bob. A shared secure random key that meets the desired the channel has been addressed in various QKD security
properties is produced at the end of the protocol, or the protocol aborts. proofs [145, 154]. Nevertheless, quantum information trans-
Reproduced from Ref. [129]. QKD, quantum key distribution. mission over long distances is still a long way from becoming a
reality, possibly until functional quantum repeaters are realised
[155]. Quantum repeaters require quantum memories with
modifies the state of the measured system [130]. Additionally, memory times close to a second [156]. As a result, the certi-
Eve is forbidden by the quantum non‐cloning theorem [8] to fication of the physical hardware remains a critical challenge
make a perfect copy of the state Alice sends to Bob while for quantum communication. While multi‐hop quantum net-
keeping the original copy. Any action by the eavesdropper works are capable of providing better rates at longer distances,
would lead to her detection and Alice and Bob can abort the they have limited achievable key rates [157].
protocol and discard the keys. The classical channel is Compared to DV‐QKD, the class of CV‐QKD protocols
authenticated such that the eavesdropper can only listen to have gained interest because of their potential technology
communication about the basis Alice and Bob used to measure advantage such as offering an improved secret key rate and
the quantum states but not the actual measurement results. being much more compatible with the existing standard optical
Therefore, Eve cannot recover information about generated fibre communication system [17, 146, 158, 159]. The CV‐
secret keys. As a result, combining a QKD protocol with a QKD schemes depend heavily on coherent detection and
secure one‐time pad encryption algorithm promises digital signal processing techniques, which are closely related to
information‐theoretic or unconditional security [11, 131]. classical coherent communication systems. Despite the signif-
Since Bennett and Brassard's (BB84) [16] and Ekert 1991 icant differences between classical and quantum optical
(E91) [126] now represent protocols using QKD, various communication systems, the knowledge gained through clas-
QKD protocols and schemes have been developed to over- sical coherent communication systems over the past few years
come the mismatch between the theoretical security of the can be applied to quantum optical communication systems,
communication model and the security of practical imple- making CV‐QKD implementations possible. However, due to
mentations [132–134]. These protocols belong to a class of imperfect devices and insufficient assumptions, a practical CV‐
discrete‐variable (DV‐QKD) protocols because the informa- QKD system may still be vulnerable to various quantum
tion is encoded in a discrete manner by using single photons to hacking attacks [160]. Most significantly, these challenges limit
carry information, exploiting, for instance, different polar- practical QKD security.
isation directions to realise the two states of a qubit [3, 9]. Considering the difficulty of calibrating the measurement
Another approach to QKD is the continuous‐variable (CV‐ device [161], an adversary could manipulate the measurement
QKD), in which information is encoded in the properties of device by sending unexpected signals [162–164]. The MDI‐
light that are continuous, for instance, the quadrature com- QKD schemes have been developed to address detection
ponents of the electromagnetic field, yielding continuous loopholes in QKD systems [165], and TF‐QKD protocols
values as measurement results [135] and the distributed‐phase‐ have been proposed for long‐distance communication with
reference protocols in which information is encoded in time QKD keys [166]. Although improved security can be achieved
and phase of weak coherent pulses [3, 17]. Owing to the de- by selecting optimal QKD resources, this requires time‐
mand for various practical implementation requirements has consuming algorithms, such as exhaustive traversal or local
led to developments such as decoy‐state QKD [136], device‐ searches [167–169]. Theoretically, as aforementioned, CV‐
independent QKD or measurement‐device independent QKD protocols lack symmetry, and security proofs involving
QKD (MDI‐QKD) [137] as well as Twin‐Field QKD (TF‐ discrete modulation rely mainly on numerical methods [150,
QKD) [138] protocols. Detailed descriptions of these ap- 170]. Although numerical methods offer tight security bounds
proaches, including proofs of their security, may be found in for various QKD protocols [149], their high implementation
Refs. [3, 9, 17, 139]. As a result of these developments, there requirements remain a challenge [170–172] as they require large
can be no doubt that QKD has emerged to be the most amounts of computational resources and time. Furthermore,
advanced practical quantum communication technology [129]. numerical methods require minimising a convex function over
 26328925, 2024, 3, Downloaded from [Link] by Pondicherry University, Wiley Online Library on [10/08/2025]. See the Terms and Conditions ([Link] on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
206
- MAFU

all eavesdropping attacks related to the experimental data [173] the choice of intensities (i.e. signal states and decoy states) and
and depend on numerous QKD parameters. Unfortunately, probabilities of sending them, which is an essential step in
this may require a significant amount of time for the corre- attaining optimal performance [178], especially when one
sponding optimisation [150, 170]. Due to these challenges, it is realistically considers statistical fluctuations due to finite‐size
imperative to develop techniques that are more useful than key in practical communications [179, 180]. Most often,
standard numerical modelling methods. exhaustive search or local search is used to optimise parameters
ML and AI techniques provide powerful tools for learning in point‐to‐point QKD systems. Although these methods are
from unknown features or complex patterns [94]. They feasible for small‐scale QKD networks in the future, they are
demonstrate significant potential in predicting optimisation not yet suitable for large‐scale networks. Accordingly, in the
parameters from noisy measurement data and address chal- following section, we examine how various AI and ML models
lenges in practical quantum communications [174]. Particularly, have been integrated to enhance QKD performance.
AI and ML models have been developed for optimising various In 2020, Ding et al. [181] used random forest (RF) instead
system parameters, such as equipment parameters or mea- of traditional search algorithms to directly predict optimal
surement device parameters [175], physical parameters of sig- parameters based on any given system conditions for finite
nals [176], and optimising the raw secret key [177]. data for MDI‐QKD and MDI‐BB84 QKD protocols. An RF
Fundamentally, the integration of AI and ML in quantum model is particularly useful for various ML tasks such as
communication systems addresses both QKD characteristics classification and regression. This is due to its high prediction
(e.g. key rate performance and practical security) as well as accuracy, tolerance to outliers and noise, and low overfitting
system components (e.g. source, channel, and measurement probability [93]. This method waives the simulation and iter-
devices). This has resulted in improved QKD schemes and ation methods, which are inherent in traditional search tech-
long‐distance transmissions. Parameter optimisation refers to niques. The results of the RF model are shown in Figure 3.

F I G U R E 3 (a) An illustration of the RF

model, where D denotes the original training
set, Dk(k = 1, …, K) denotes the subset
sampled with replacement from D, and fk
denotes a decision tree grown on Dk.
(b) Comparison of secure key rate between
RF and LSA for MDI QKD protocol. The
average ratio is 0.9941 and the standard
deviation is 0.0207. (c) Comparison of secure
key rate between RF and LSA for MDI‐BB84
QKD protocol. The average ratio is 0.9944
and the standard deviation is 0.1438.
(d) Typical search space of input features: e,
d, N and L for fibre loss α = 0.2 dB/km,
detector efficiency ηd = 14.5%, and error
correction efficiency fe = 1.16.
(e) Comparison of prediction accuracy when
using the RF and NN model and a better
generalisation is achieved in each situation.
Reproduced from Ref. [181]. LSA, local
search algorithm; MDI QKD, measurement‐
device independent quantum key
distribution; NN, neural network; RF,
random forest.
 26328925, 2024, 3, Downloaded from [Link] by Pondicherry University, Wiley Online Library on [10/08/2025]. See the Terms and Conditions ([Link] on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
MAFU
- 207

Specifically, the RF model assumes the training set considered. Using these ML models, predictive parameters
D ¼ fXi ; Yi gni¼1 , where Xi are input features with responses significantly reduce optimisation time while maintaining high
Yi. As part of the RF model, bootstrap aggregating or bagging accuracy. Notably, XGBoost's performance is better than
is applied repeatedly (i.e. K times) to randomly n examples, BPNN and RF in parameter prediction. This approach elimi-
with their replacement from the training set with an inde- nates the need for simulation and iterations in the search
pendent bootstrap sample, Dk. Trees are grown based on Dk, method which requires real‐time optimisation of the future
with each node selecting a random subset of features and the QKD network.
best split being determined from those features. After training, Practical QKD implementations require efficient, real‐time
the predictors for a fresh sample X0 can be obtained by feedback control to maintain stability if there are disturbances
P
averaging the predictions from all tress f ¼ K1 Kk¼1 f k ðX 0 Þ, from either the external environment or imperfect compo-
where fk(X0 ) is the prediction on X0 of an individual tree. For nents within the system. The ‘scanning‐and‐transmitting’
the 3‐intensity MDI‐QKD protocol, Alice (Bob) should programme is used to compensate for physical parameter
optimise the intensities of signal state and decoy state (μ, v), variations of the devices, which can provide accurate
and the probability of intensities and bases (Pμ, Pv, PX|μ, PX| compensation, but may require a considerable amount of time
to stop and calibrate the processes, which may reduce the
v), where Pμ(PX|v) is the probability to choose signal (decoy)
state and PX|μ(PX|v) denotes the probability to choose the X efficiency of key transmission. As a result, Liu et al. [183]
basis conditional on signal (decoy) state. These parameters are proposed a practical phase‐modulation in QKD that employs
grouped as vectors ! v , which represent responses. Note that an ML model, specifically the LSTM network to predict
physical parameters of devices in advance and actively perform
the key rate is influenced by the system's conditions ! s,
real‐time control on corresponding QKD devices. Figure 4
including the misalignment ed, the dark count rate d, the de-
shows the schematic setup and results of the experimental
tector efficiency ηd, the pulse number N, and the transmission
study. This experiment runs the BB84 QKD system by
distance L between Alice�and Bob. � The key rate is calculated applying either the traditional ‘scanning‐and‐transmitting’
! !
using the equation r ¼ R s ; v . Optimal secure key rate is programme or the proposed LSTM model. During trans-
achieved searching !
by v opt according to ! v opt ðsÞ¼ mission, the 3‐intensity (μ = 0.5, v = 0.1, 0) decoy‐state
� � method is implemented and the corresponding quantum bit
argmax!v ∈V
R s ; v , where V is the search space of !
! ! v. error rate (QBER) and counting rates for 2 days are recorded.
! !
Since v is a function of the given system condition s it is The results are shown in Figure 4a–c. Notably, the ‘duty ratio’
opt
challenging to simulate analytically. Particularly, using the RF of the traditional ‘scanning‐and‐transmitting’ programme and
model results in reduced time and hardware overhead the proposed LSTM model‐based system is 50% and 83%,
compared to traditional search methods. Moreover, this model respectively. As a demonstration that the proposed LSTM has
achieves a high prediction accuracy of over 99% of the optimal the ability to predict continuously over an extended period, the
secure key rate. This holds significant promise for future QKD proposed LSTM model‐based QKD system was run for
applications. 10 days at a distance of 50 km, and the results are shown in
In MDI‐QKD protocols, selecting parameters, such as the Figure 4d. Specifically, the results demonstrate no trend of
probability of selecting an X‐basis or Z‐basis, the signal state deterioration from start to finish, demonstrating the long‐term
intensity and decoy state, as well as system calibration, be- reliability and stability of the proposed ML model. According
comes more challenging as the number of parties in the to these authors, this ML model is a promising candidate for
network increases. The most common method for optimising large‐scale quantum communication networks.
parameters at the moment is to use optimisation algorithms. The implementation and performance of CV‐QKD are
Even though the method is accurate, it is time‐consuming and threatened by various attack strategies. To counter these at-
consumes hardware resources. Rather than searching for tacks, various real‐time monitoring modules are exploited to
optimised parameters, Lu et al. [182] present a novel back prevent different types of attacks. Due to the uncertainty
propagation neural network (BPNN) capable of predicting associated with the estimation of excess noise, this strategy
optimised parameters with fewer resources and at a higher lacks a universal defence method, which prompted Mao et al.
speed. Moreover, it is notable that the BPNN addresses the [184] to propose a defence strategy to address these disad-
challenge of system recalibration, which is experienced when vantages and resist the majority of known attacks on CV‐QKD
working with large‐scale MDI‐QKD networks. The BPNN systems. By analysing several pulse characteristics that are
can be used to address system recalibration in real‐time. This is affected by a variety of types of attacks, the authors deduce a
achieved by requiring the use of some discarded data generated feature vector that is used as input to an ANN model for
by the communication process rather than the addition of analysis. As a result of the trained ANN model, attacks can be
additional devices or a complete scan of the system. automatically identified and classified with a precision and
On the other hand, Dong et al. [174] use the extreme recall rate exceeding 99%. By implementing most of the
gradient boosting (XGBoost), BPNN, and RF model to pre- known attack strategies, the proposed scheme significantly
dict the TF‐QKD optimisation parameters affecting the key increases system security. However, compared to a CV‐QKD
rate and transmission distance. While a TF‐QKD, overcomes system that did not adopt any countermeasures against at-
the basic limits of QKD without repeaters, in practice, it still tacks, the proposed scheme slightly decreased the transmission
needs to optimise all parameters when a finite data size is distance and secret key rate.
 26328925, 2024, 3, Downloaded from [Link] by Pondicherry University, Wiley Online Library on [10/08/2025]. See the Terms and Conditions ([Link] on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
208
- MAFU

F I G U R E 4 Schematic of the experimental setup. BS, beam splitter; CB, control board; FM, Faraday mirror; FMI, Faraday‐Michelson Interferometer; OPM,
optical power metreer; PM, phase modulator; SPD, single‐photon detector; THD, temperature, and humidity detector. (a, b) Comparisons between applying the
traditional ‘scanning‐and‐transmitting’ programme and using present LSTM model‐based QKD systems with respect to time and average values at the
transmission distance of 50 and 150 km. Notably, the QBER of the traditional method is linked to the left Y‐axis, and of LSTM model is linked to the right Y‐
axis. (c) The key generation rate versus transmission distance. In each figure, the square points refer to the results of applying the traditional ‘scanning‐and‐
transmitting’ programme, and the circle points correspond to using our new method. (d) Comparisons between applying traditional scanning‐and‐transmitting
programme and using current LSTM‐model‐based QKD systems on QBER at the transmission distance of 100 km. Reproduced from Ref. [183]. LSTM, long
short‐term memory; QBER, quantum bit error rate; QKD, quantum key distribution.

Another important approach to QKD practical systems is communication failures contributes to improved practical
underwater CV‐QKD. However, the implementation of un- implementation feasibility of CV‐QKD. Due to the vulnera-
derwater QKD faces challenges such as such as equipment bility of CV‐QKD systems to various quantum hacking attacks,
imperfections, channel disturbances, and eavesdroppers, which Ding et al. proposed an ML‐based attack detection scheme
may cause invalid communication. As a result, Li et al. [56] (MADS) as a universal defence strategy [160]. According to the
investigate the risk ensemble learning (EL) approach caused by results, the MADS can detect most quantum hacking attacks
channel disturbance resulting in communication failure. To and revise overestimated secret key rates resulting from a CV‐
improve the security of the CV‐QKD system with discrete QKD system that does not employ a defence strategy to
modulation, the authors present an EL approach with a self‐ achieve a tighter security bound.
adaptive structure and predictability of communication fail- Towards addressing measurement device or detection
ures. Based on the EL approach, the current communication challenges, using an ML‐assisted MDI‐QKD system, Zhang
status of the discrete‐modulated CV‐QKD (DM‐CV‐QKD) et al. [185] demonstrate how phase drift between two users can
system can be predicted based on Bob's phase data, enabling be predicted in advance and compensated actively in real‐time,
adaptive judgement of the current communication status, and dramatically increasing key transmission efficiency. Specifically,
ensuring typical QKD communications. As opposed to the they demonstrate the application of the LSTM model to the
traditional ML process used for CV‐QKD systems, this MDI‐QKD system for calibrations of reference frames. This
method does not require additional parameter detection schematic setup and results are shown in Figure 5. By utilising
equipment. This means the scheme is not susceptible to se- this model, the QKD system can predict phase drift between
curity vulnerabilities caused by parameter detection equipment. two users in advance, and compensate for it in real‐time. The
Moreover, the EL approach can reach 70%, which is higher MDI‐QKD scheme generally achieves lower secret key rates
than other learners, especially in logistic regression, support than the BB84 QKD scheme, while achieving a balance and
vector machines, and BPNNs. As a result, this ability to predict practicality with finite data size effects. As a means of
 26328925, 2024, 3, Downloaded from [Link] by Pondicherry University, Wiley Online Library on [10/08/2025]. See the Terms and Conditions ([Link] on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
MAFU
- 209

F I G U R E 5 (a) Schematic setup of the MDI‐QKD system. ATT, attenuator; BS, beam splitter; EPC, electronic polarization polarisation controller; FM,
Faraday mirror; IM, intensity modulator; Laser, continuous‐wave laser; PM, phase modulator; SNSPD, super‐conducting nanowire single‐photon detector.
(b) Theoretical and experimental results of the key rate's dependence on the transmission distance. The system parameters are as follows: dark counting rate per
pulse and overall efficiency of detection side are 4 � 10−8 and 60%; misalignment errors in Z and X bases are 0.15% and 1.5%, respectively. (c) Comparisons
between applying traditional scanning‐and‐transmitting programme and using current LSTM‐model‐based QKD systems on QBER at the transmission distance
of 100 km. Reproduced from Ref. [185]. LSTM, long short‐term memory; MDI‐QKD, measurement‐device independent QKD; QBER, quantum bit error rate;
QKD, quantum key distribution.

increasing the key rate, the authors demonstrate a variation of Furthermore, the method avoids potential vulnerabilities since
the LSTM model based on a simple three‐intensity decoy‐state no additional quantum resources are required. The programme
MDI‐QKD scheme with biased basis choice, in which the optimisation can further improve the duty cycle and stability
decoy pulses are only prepared in the X‐basis. This protocol time to match the requirements of different systems.
consists of signal pulses prepared either on a Z or X basis, Furthermore, the proposed system maintains excellent stability
which is distinct from the traditional three‐intensity decoy state over a long period, conclusively demonstrating its effective-
schemes [186]. However, similar to the approach demonstrated ness. As a final point, this approach can serve as a valuable
in Ref. [183], authorised parties randomly prepare phase‐ reference when it comes to the implementation of future large‐
randomised weak coherent state pulses in three different in- scale quantum communications.
tensities (μ, v, 0) with certain probability values that corre- Notably, Ismail et al. demonstrated using the EL technique
spond to the intensity of the signal state, the decoy state, and called RF to predict the strength of the atmospheric turbulence
the vacuum state. The experimental demonstration demon- in the quantum channel in free space using the Strehl Ratio [187].
strated improved secure key exchange for single‐mode fibres The authors estimate the Strehl ratio of the quantum channel
over 100 and 250 km. By utilising this approach, MDI‐QKD with a mean absolute percentage error of 4.44%. The authors
systems show a significant improvement in transmission effi- also perform predictions using linear and support vector re-
ciency over conventional scanning‐and‐transmitting systems. gressors; however, the RF performs better. As part of the
 26328925, 2024, 3, Downloaded from [Link] by Pondicherry University, Wiley Online Library on [10/08/2025]. See the Terms and Conditions ([Link] on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
210
- MAFU

process of characterising quantum channels, these values pro- devices and channels for four examples: MDI‐QKD, BB84,
vide insight into the level of disturbances that must be overcome and TF‐QKD. According to the authors, the approach is
during key distribution. This allows one to accurately charac- general and does not depend on any particular protocol. With a
terise quantum channels and assist in mitigating the challenges fraction of the power consumption of local search, this tech-
related to information losses in transmission channels. Hence, nique is suitable for low‐power devices and can achieve 2–3
using ML to monitor free‐space quantum channels will enable orders of magnitude faster optimisation speed than local
real‐time monitoring of QKD systems. In 2022, Lollie et al. [188] search. The programme can be run on either NN acceleration
demonstrated high‐dimensional encryption in optical fibres us- chips or on common CPUs with relatively low performance,
ing spatial modes of light and ANNs. A particular feature of this depending on their preference. Besides this being extremely
communication protocol is that it allows the recovery of mes- useful for free‐space QKD applications that require low la-
sages and images with a high level of accuracy. This protocol tency and low power budgets, it is useful for quantum internet‐
provides a way to increase the quantum information capacity per of‐things (IoT). The IoTs allow even a small portable device
photon while maintaining the fidelity of information transfer connected to numerous users to optimise everything parame-
that is essential for quantum technologies utilising structured ters in real‐time. Furthermore, this simple demonstration
fields of light, particularly those whose propagation is challenged suggests that similar methods may apply to other optimisation
by free space. Moreover, in 2018, Ou et al. [189] developed a tasks commonly encountered in the design and control of
method to assist coexistence between quantum and classical practical QKD systems, including the determination of the
communication channels. They predicted the quantum channel optimal post‐selection thresholds in free‐space QKD and the
QKD quality in a dense wavelength‐division multiplexing tuning of polarisation controller motors to control misalign-
(DWDM) fibre network with multiple ML models. The authors ment. This could be useful for reference‐free independent
demonstrate that the K‐neighbours regression is the most ac- QKD schemes [191].
curate model to spectrally re‐allocate the channels for stable Moreover, in 2018, Li et al. [177] presented a distance‐
performance. The feasibility of the ML application was weighted K‐nearest neighbours (DW‐KNN) algorithm to a
confirmed by several field trials, which demonstrated robust key DM‐CV‐QKD to preprocess the raw key data before per-
generation and distribution. Furthermore, Liu et al. [176] forming the reconciliation. This improves the accuracy of the
demonstrated the advantages of using a support vector regres- raw key data and the performance of DM‐CV‐QKD. The pro-
sion model for optimising the performance and security of fibre posed scheme could be employed to overcome several impair-
QKD systems. This SVR‐based approach provides optimal ments induced by the channel, thereby lowering the demand for
performance and practical security for the QKD system. As the error correction codes on the signal‐to‐noise‐ratio threshold of
system does not require additional resources and no real‐time the quantum channel. Classical error correction techniques are
monitoring module, it supports automatic prediction of signal used to detect and correct errors in the transmission of classical
physical parameters over time and can be applied to any signal information [3, 9]. This technique plays a crucial role in
physical parameter that can be measured in a practical QKD improving the security and reliability of QKD systems. Niemiec
system. It is noteworthy that Okey et al. [169] showed that the demonstrates a key reconciliation technique that uses ANNs
Tree‐CNN algorithm is an efficient method for selecting the (TPM machines) to reconcile errors in quantum channels [78].
optimal QKD (i.e. BB84 for short distances and TF‐QKD for The authors argue that error correction using ANNs is robust to
long distances) for the implementation of large‐scale commu- current attacks imposed by quantum computers. Furthermore,
nication systems in real‐time. The proposed model is compared Long et al. [81] presents an overview of how ML has been applied
to other ML techniques based on convolutional neural networks to almost every stage of the QKD protocol, particularly CV‐
(CNN). This study showed a 99.89% Area Under Curve during QKD. The study indicates that ML has been used in phase er-
testing and 0.65 s time‐cost performance, which outperformed ror estimation, excess noise estimation, state discrimination,
the results obtained in related studies in predicting the best QKD parameter estimation and optimisation, key sifting, information
protocol. The proposed scheme was further validated using reconciliation, and key rate estimation. To improve time and
different transmission distances and three QKD protocols to resource consumption in CV‐QKD, Liu et al. [171] improve a
demonstrate that the predictions and actual results matched one NN model predicting key rates in nearly real‐time by combining
another. The proposed model demonstrated a fast, reliable, and the model with Bayesian optimisation. The combined model
precise solution to the problem of selecting an appropriate QKD automatically designs the best architecture of neural networks
protocol based on performance and time. The authors suggest computing the key rates for DM‐CV‐QKD protocols in real‐
that long distances established using TF‐QKD implementations time. Precisely, two variants of CV‐QKD protocols with qua-
may eliminate the need for quantum repeaters in certain in- ternary modulation were studied using the model. The results
stances. Particularly, other ML models have also been applied to indicate high reliability with a secure probability as high as
fibre‐based optical communication systems [101, 190] as well as 99.15%–99.59%, tight secure bounds, and high efficiency with a
nanoscale devices characterisation and parameter estimation, speed‐up of approximately 107 than the numerical methods. A
leading to improvements in quantum communication sys- key benefit of the model is that it allows real‐time automatic and
tems [125]. efficient computation of QKD. This will address the growing
A NN is presented by Wang and Lo [178] in 2019 which demand for implementing QKD protocols on mobile platforms
predicts the optimal parameters based on the characteristics of such as handheld QKD systems, drone‐based QKD, or satellite‐
 26328925, 2024, 3, Downloaded from [Link] by Pondicherry University, Wiley Online Library on [10/08/2025]. See the Terms and Conditions ([Link] on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
MAFU
- 211

ground QKD. The authors argue that using neural networks to state without transmitting the encoded particle itself [24, 196],
optimal parameters in their work outperforms results in Refs despite the quantum no‐cloning theorem [8]. This phenome-
[176, 178, 181, 182]. non is influenced by the fundamental principle of quantum
In 2020, Liao et al. [192] proposed a multi‐label learning‐ entanglement and therefore cannot be simulated with classical
based scheme for discretely‐modulated CV‐QKD (ML‐CV‐ channels [63, 197, 198]. This enables it to serve as a valuable
QKD). This scheme divides the quantum system into a trusted resource for overcoming distance limitations in quantum
state learning process for training and estimating classifier, and communication and quantum networks, as well as the difficulty
an untrusted state prediction process for generating the final of establishing long‐range interactions among qubits in quan-
secret key. To better represent the characteristics of modulated tum computation [199, 200]. An overview of the teleportation
coherent states, feature extraction was used. This study also scheme based on the Bennett et al. [24], is illustrated in
investigates how the multi‐label classification algorithm Figure 6. The quantum state teleportation scheme consists of
embedded in ML‐CV‐QKD can be used for the prediction of an unknown quantum state being transmitted from the sender,
unknown signals. The results illustrate that the MLCA‐ Alice to the receiver, Bob. The two authorised participants are
embedded ML‐CV‐QKD is both feasible and effective at connected by a quantum channel and a classical channel. The
predicting unknown signals. The numerical simulations show quantum channel consists of an entangled pair shared between
that the proposed MLCA‐embedded ML‐CV‐QKD out- Alice and Bob:
performs existing CV‐QKD protocols, specifically in terms of
maximum transmission distance, and the performance of both 1
jψ − 〉23 ¼ pffiffi ðj0〉2 j1〉3 − j1〉2 j0〉3 Þ; ð1Þ
transmission distance and secret key rate will continue to in- 2
crease with the increase of modulation variance. Moreover,
Zhou et al. [172] developed an NN that can be used to predict
the information‐theoretically secure key rates of homodyne where particles 2 and 3 are held by Alice and Bob, respectively.
detection DM‐CV‐QKD with high probabilities of up to This shared entangled state is one of the four maximally
99.2% at distances of 0–100 km with <0.015 excess noise. entangled Bell states jΦ� 〉 ¼ p1ffi2ffi ðj0〉j0〉 � j1〉j1〉Þ and
According to the authors, their method achieves a speed that is jΨ� 〉 ¼ p1ffi2ffi ðj0〉j1〉 � j1〉j0〉Þ. Notably, Charlie provides the
six orders of magnitude higher than the numerical method in input particle 1 to be teleported to Bob in a general quantum
Ref. [149]. Since it takes a certain time for the QKD system to state:
collect data, the predictability of the key rates by the NN meets
all practical requirements. This technique can in principle be jψ A 〉 ¼ αj0〉1 þ βj1〉1 ; ð2Þ
applied to any protocol with reliable numerical methods.
Furthermore, there is also increasing interest in applying ML where the complex coefficients α and β satisfy |α|2 þ |β|2 = 1.
techniques to suppress noise [60, 171, 193]. For an extended Alice performs a Bell‐state measurement (BSM) and randomly
discussion on enhancing various CV‐QKD systems with ML, projects particles 1 and 2 each to one of the four Bell states with
including detection methods and attack methods, we refer the equal probability. Finally, Alice informs Bob of the BSM result
reader to Refs. [78, 176, 177, 194]. These works demonstrate through the classical channel, and Bob performs the corre-
that by strategically combining quantum mechanics and sponding Pauli or combinations of operators {I, X, Z, ZX} on
advanced computational algorithms, various AI and ML his particle according to the results to recover the unknown
techniques can be used to optimise key generation rates, reduce quantum state |ψA〉 of the particle of Charlie. Notably, the same
error rates, and increase the reliability and transmission dis- scheme can serve for entanglement swapping when the qubit to
tances of QKD systems. As QKD technology matures, be teleported is itself entangled with a fourth qubit held by
standardisation efforts are underway to ensure its sustainability another party.
and reliability soon. Standardisation is critical to accommodate Since the Bennett et al. proposal in 1993 [24], quantum
a variety of potential use cases and several plausible QKD teleportation has been demonstrated using various platforms
protocols. Regarding the progress in standardisation and the and technologies, including single atoms [201, 202], BSM
practical challenges that prevent the widespread implementa- [203], photonic chips [14], time‐bins [204, 205], atomic en-
tion of QKD in our future communication networks, we refer sembles [206], solid‐state quantum systems [207, 208], nuclear
the reader to Refs. [29, 129]. The Table 1 provides a summary magnetic resonance [209], quantum optics [195, 196, 210–
of the AI/ML solutions proposed for each QKD scheme. 215], spins [199], photons [216], and superconducting qubits
Moreover, it provides an overview of the main objectives or trapped ions [217, 218], which have become integral to
outcomes of each application. numerous quantum computing architectures [200, 219], where
it allows quantum information to be ‘swapped’ between
different qubits. In awe of these developments, recent theo-
3.2 | Quantum teleportation retical and practical advances in quantum teleportation focus
on understanding its nonclassical nature and its potential ap-
Quantum teleportation remains one of the most significant plications. In terms of practical applications, outstanding
protocols in quantum communication [5, 24, 195]. It enables teleportation distances of 100‐km [220], 1200‐km [221], and
the nonlocal transmission of an arbitrary unknown quantum 1400‐km [222] using free‐space optical links as well as 100‐km
 26328925, 2024, 3, Downloaded from [Link] by Pondicherry University, Wiley Online Library on [10/08/2025]. See the Terms and Conditions ([Link] on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
212
- MAFU

TABLE 1 Summary of AI/ML applications in QKD schemes.

QKD protocol or device AI/ML model Main objective or outcome

MDI‐QKD BPNN Predicted various optimised parameters using limited resources [182]

MDI‐QKD LSTM Addressed measurement device or detection challenges and predicted in advance
phases drifts between two users and compensates for it in real‐time [185]

BB84 & MDI‐QKD LSTM Predicted physical parameters in advance and actively performed real‐time control of
corresponding QKD devices [183]

MDI‐QKD & MDI‐BB84‐QKD RF Predicted optimal parameters based on any given system conditions for finite
data [181]

TF‐QKD XGBoost/BPNN/RF Predicted the TF‐QKD optimisation parameters influencing the key rate and
transmission distance. Eliminated the need for simulation and iterations in the
search method. Results of XGBoost outperform the BPNN and RF [174]

BB84 & TF‐QKD Tree‐CNN Accurately selected optimal QKD parameters [169]

MDI‐QKD, BB84 QKD and Neural network Accurately and efficiently predicted the optimal parameters [178]
TF‐QKD

CV‐QKD LSTM Enhanced the performance and stability of the protocol [185]

CV‐QKD BPNN Adjusted the modulation variance to ensure an optimal system with a higher
achievable key rate under varied atmospheric turbulence intensities [54]

CV‐QKD ANN Detected quantum attacks in the communication system [184]

CV‐QKD Ensemble learning Predicted channel failure due to disturbance in underwater communication
systems [56]

CV‐QKD MADS Detected most quantum hacking attacks and revised overestimated secret key
rates [160]

CV‐QKD Support vector regression Optimised the performance and the practical security of a fibre QKD system [176]

CV‐QKD ANN Experimentally addressed a quantum attack defense strategy [184]

DM‐CV‐QKD KNN Improved the accuracy of the raw key data and secure distance of DM‐CV‐QKD
systems [177]

DM‐CV‐QKD Neural network þ Bayesian Computed key rates in real‐time on a low power platform [171]
optimisation

DM‐CV‐QKD Neural network Predicted information‐theoretically secure key rates of homodyne detection DM‐
CV‐QKD with a great probability (up to 99%) at a distance of 0–100 km and an
excess noise of no more than 0.015 [172]

Quantum channel RF Predicted the atmospheric strength of quantum channels in free space [187]

Quantum channel KNN regression Predicted the quantum channel quality in a DWDM fibre network [189]

Optical fibre ANN Recovered messages and images with a high level of accuracy [188]

Key distillation ANN Corrected errors occurring during transmission [78]

Abbreviations: AI, artificial intelligence; ANN, artificial neural network; BPNN, back propagation neural network; CNN, convolutional neural network; CV‐QKD, continuous‐variable
QKD; DM‐CV‐QKD, discrete‐modulated CV‐QKD; KNN, K‐nearest neighbour; LSTM, long short‐term memory; MADS, ML‐based attack detection scheme; MDI‐QKD,
measurement‐device independent QKD; ML, machine learning; QKD, quantum key distribution; RF, random forest; TF‐QKD, twin‐field QKD.

[214] and 143‐km [223] via commercial optical fibre channels the theoretical model. These include entanglement swapping
are observed. These performances signify a key step towards [229–231], controlled quantum teleportation [232, 233], tele-
realising quantum communication at a global scale [22, 199, portation of a shared quantum secret [234, 235], quantum
220, 224–226] and establish a promising avenue towards a teleportation network [236], port‐based teleportation [237,
quantum internet [23, 213, 227, 228]. Thus, quantum tele- 238], quantum gate teleportation, and quantum computing
portation over long distances is essential for realising global [239, 240].
quantum communications as well as large‐scale quantum Most significantly, various ML techniques have been
networks. Over the years, several quantum teleportation applied in detecting or classifying quantum entanglement, a
models have been widely investigated theoretically and valuable resource in quantum teleportation, in various settings.
experimentally (see Refs [63, 199]), including quantum tech- For instance, in 2018 Lu et al. [241] demonstrated a reliable
nology protocols and others that are conceptually valuable in classifier for classifying entangled and non‐entangled states by
 26328925, 2024, 3, Downloaded from [Link] by Pondicherry University, Wiley Online Library on [10/08/2025]. See the Terms and Conditions ([Link] on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
MAFU
- 213

be coded to create a quantum wormhole with the Google

Sycamore quantum processor [243]. Through quantum infor-
mation passing through the system, the researchers sought
insights into gravitational dynamics. Most significantly, these
examples demonstrate that this field is advancing, and inte-
grating AI and ML algorithms promises significant potential to
optimise quantum state control and manipulation, which is
critical in teleportation. Broadly, this could facilitate the dis-
covery of novel teleportation techniques and enable long‐
distance and high‐fidelity quantum communication.
The research on quantum teleportation has demonstrated
significant progress, suggesting quantum teleportation of
complex quantum states (i.e. multiple degrees of freedom
[216], high‐dimensional quantum states [244, 245]) with high
F I G U R E 6 An illustration of a quantum state teleportation scheme. fidelity under the premise of long distances [222] is possible in
The scheme consists of previously shared entanglement, BSM, classical
these systems in the near future. However, since quantum
communication, and unitary operations. The red, yellow, and blue circles
represent particles in the quantum teleportation protocol. The SA, SB, SC teleportation inherently depends on manipulating complex
represent the quantum channels for transmitting particles A, B and C. The quantum states, some inherent challenges remain. These
states |ψA〉 and |ψB〉 represent the input quantum state and output challenges are associated with the preparation of entanglement
quantum state, respectively. The EPR pair is a maximally entangled [199, 246, 247], Bell‐state measurements [248], the randomness
quantum state. Reproduced from Ref. [63]. BSM, Bell‐state measurement;
of the spontaneous parametric down‐conversion [249] and
EPR, Einstein–Podolsky–Rosen.
inefficient nonlinear processes [250]. Most significantly, efforts
such as implementing quantum teleportation via a determin-
employing supervised learning techniques. Specifically, the istic photon source [251, 252], highly efficient on‐chip
authors developed a reliable separability‐entanglement classi- nonlinearity [253], high‐dimensional entanglement of
fier by combining supervised learning and the convex hull trapped‐ion [254], nitrogen‐vacancy (NV) centres [255] and
approximation method. This method outperforms conven- superconducting qubits [256] have been poised as promising
tional criteria for entanglement detection in generic cases in solutions [63].
terms of speed and accuracy. Furthermore, the authors claim Due to some of these outstanding challenges, several ML
that the classifier can be extended to higher dimensions in the algorithms have recently been proposed to enhance and opti-
future and incorporated into experiments involving entangle- mise quantum teleportation protocols. A study conducted in
ment engineering. In 2021 Chen et al. [64] proposed a pro- 2019 by Wallnöfer et al. [55] develops the mathematical tools
cedure for detecting quantum entanglement with an needed to perform reinforcement learning of central quantum
unsupervised ML method (complex‐valued NN) for high‐ communication protocols, specifically quantum teleportation,
dimensional and multipartite quantum systems. A noteworthy entanglement swapping, and entanglement purification. This
feature of this NN is that it achieves a detection accuracy of was achieved by combining reinforcement learning techniques
97.5%. This work is critical since quantum features, such as with projective simulations. Particularly, a projective simulation
entanglement and coherence, are indispensable resources in is a physically motivated framework for reinforcement learning
various quantum information processing tasks, such as QKD, and decision‐making, based on deliberation in an episodic and
teleportation, and quantum computing [5]. Additionally, by compositional memory (ECM) [55]. An example of this is
harnessing ML algorithms to detect entanglement, recently in Figure 7, which illustrates the trial‐and‐error process of
2023, Asif et al. [242] use supervised ML methods to construct learning quantum communication protocols through rein-
a classifier by employing the relationship between coherence forcement learning interactions between an agent and the
and entanglement. In particular, this method encodes multiple environment. An agent is equipped with a universal gate set,
Bell‐type inequalities (as an entanglement witness) for the and it specifies the desired task through a reward scheme. By
relative entropy of coherence into ANNs for detecting trial and error, the agent manipulates quantum states to create
entangled and separable states in a quantum dataset. Notably, quantum communication protocols. When the agent (the
the classifier trained with the most straightforward NN dis- protocol) interacts with the protocol (the environment), the
tinguishes the quantum states with an accuracy of 78.18%, and agent perceives the present state of the protocol (the envi-
the classifier's performance increased by increasing the number ronment) and selects one of its available options (actions).
of neurons to result in a performance of 94.62%. As a result, Thus, the previous protocol is revised and the interaction step
this brief review illustrates the potential of using ML tech- is completed. A reward function specifies feedback given to
niques to detect entanglement thereby enhancing quantum agents during each interaction step. This is based on the spe-
communication protocols, including teleportation. cific quantum communication task (a–d) illustrated in Figure 7.
Furthermore, another important but different aspect, Notably, a reinforcement learning agent interprets a reward and
Boerkamp (2023) demonstrates the utility of ML to calculate updates its memory accordingly. The agent plays a more critical
the minimum number of qubits required and how they should role than mere parameter estimation since simple search
 26328925, 2024, 3, Downloaded from [Link] by Pondicherry University, Wiley Online Library on [10/08/2025]. See the Terms and Conditions ([Link] on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
214
- MAFU

F I G U R E 7 An illustration of a reinforcement‐learning agent interacting with the environment. The agent performs actions that change the state of the
environment, while the environment communicates information about its state to the agent. The reward function is customised for each environment. ECM,
episodic and compositional memory; PS, projective simulation. The initial states for the different environments that we consider here are illustrated:
(a) teleportation of an unknown state, (b) entanglement purification applied recurrently, (c) quantum repeater with entanglement purification and entanglement
swapping, (d) scaling of quantum‐repeater concepts to distribute long‐distance entanglement. Reproduced from Ref. [55].

Clifford (magenta) and universal (blue) gates. As you can see

from the individual learning curves of two agents in Figure 8,
the mean squared deviation does not reach zero. A better
explanation can be found in Figure 8b by examining the in-
dividual learning curves of two example agents: the agent does
not arrive at a single solution to this problem setup, but rather
four different solutions. As shown in Figure 8c, the learning
F I G U R E 8 An illustration of a reinforcement learning algorithm in a
curve has been averaged over 500 agents. However, a signifi-
teleportation protocol. (a) Learning curves of an ensemble of projective cant number of trials are needed for the agent to discover a
simulation agents: average number of actions performed to teleport solution due to the long action sequence. However, this illus-
unknown quantum states. (b) Two learning curves (magenta and blue) of trates that the agent can locate the quantum teleportation
two individual projective simulation agents. Four solutions of different protocol without being provided with an initial entangled state.
lengths are found by agents. (c) Learning curve in the learning setting that
is, the average number of actions performed to teleport an unknown
Thus, this study provides a demonstration of how reinforce-
quantum state. In (a and c), the curves represent an average of over 500 ment learning is a more effective means of achieving quantum
agents. The shaded areas show the mean squared deviation γ/3. This communication than optimisation techniques, particularly
deviation appears not only because of the different individual histories of quantum teleportation over long distances. A key benefit of
the agents but also because of the difference in the individual solution this development is that it opens up the possibility of using ML
lengths shown in (b). Reproduced from Ref. [55].
techniques, such as reinforcement algorithms, to design and
implement quantum networks, which do not have a wide range
of existing protocols.
methods would not be sufficient for problems of this size (i.e. Owing to difficulties associated with executing high‐
the number of possible states in teleportation tasks). According dimensional Bell‐state measurements, teleportation of high‐
to Figure 7, the ECM is based on a random walk process, dimensional quantum states remains challenging [199]. Zhang
which requires little computational effort and can be acceler- et al. [62] demonstrated experimentally in 2022 that unknown
ated via quantum walk processes, which result in a quadratic qutrits can be teleported by generating, transferring, and
speedup in deliberation time, making the projective simulation manipulating photons and training quantum encoders on sili-
model conceptually appealing. con chips. The use of quantum autoencoders can facilitate the
Particularly, the learning process is depicted in Figure 8, efficient compression of quantum data. Specifically, the authors
where the two particles 2 and 3 at the sender's station and employ unsupervised ML to train an autoencoder to train an
receiver station, respectively, are described in Equations (1) and encoder capable of teleporting a quantum state from a
(2). Figure 8a, shows the learning curve, which is the number particular d to n‐dimensions (i.e. compressing‐teleporting‐
of operations the agent applies to reach a solution in each trial. decompressing a quantum state) using integrated photonic
The authors observe that the learning curve converges to some chips for future scaling. Training the encoder involves mini-
average number of operations in both cases though the mean mising the likelihood of the trash mode being occupied while
square deviation does not go to zero. A learning curve, or the retaining all information regarding qubit modes. A lossless
number of operations the agent uses in each trial, is shown in compression occurs when the trashed state is unoccupied,
Figure 8a. According to the authors, the learning curves allowing the decoder to reconstruct the initial qutrit. Notably,
converge to the same average number of operations for both ML algorithms update the encoder parameters to achieve
 26328925, 2024, 3, Downloaded from [Link] by Pondicherry University, Wiley Online Library on [10/08/2025]. See the Terms and Conditions ([Link] on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
MAFU
- 215

teleportation fidelity of 0.894. This work lays the foundation 3.3 | Quantum secret sharing
for integrating ML techniques to develop reliable and efficient
quantum networks and quantum computations, which opens Secret sharing is an essential cryptographic protocol for
the way to the quantum internet. Satellite‐based quantum sharing highly confidential information with trusted and
communications can also assist in establishing the infrastruc- untrusted individuals. Secret sharing was independently intro-
ture for a global quantum network [22, 222]. While satellite duced by Shamir [258] and Blakley in 1979 [259] as a solution
quantum technologies have some significant potential, they to safeguard cryptographic keys. Generally, in a secret sharing
face challenges such as atmospheric turbulence, diffraction, scheme, the distributor divides a classical or quantum secret
and regular refraction. These challenges cause drift or random into multiple shares, and only the shareholders in the author-
wandering and broadening of the transmitted beam. This re- ised set can retrieve the secret when a sufficient number of
sults in fluctuations in transmittance, which complicate the shares are combined. Therefore, the shareholders in the non‐
prediction process. Therefore, there is a need to combat the authorised set are unable to leak any information or recover
effects of loss and noise in the transmission link. the shared secret [260, 261]. Various secret sharing schemes
Towards addressing challenges in satellite‐based quantum have since been realised can be classified as secret splitting
communication, Xu et al. [257] proposed an ML‐assisted [262], threshold sharing schemes [258, 259], and verifiable se-
approach to continuous‐variable quantum teleportation pre- cret sharing scheme [263]. These schemes are utilised in
diction over the satellite‐ground link. The authors employ a K‐ various cryptographic protocols [264], for instance secure
nearest neighbour (KNN) algorithm and a DT algorithm to multiparty computation (MPC) [265, 266], secure aggregations
predict the squeezing parameter r and satellite altitude H, [267], multi‐signature [268, 269], and attribute‐based encryp-
which can be derived from the traditional formula. The authors tion [270, 271]. The security of these schemes is dependent on
simulate the difference between the real value and the pre- the computational complexity of large prime decomposition.
dicted value using different zenith angles, turbulent intensities, As quantum computers and quantum algorithms develop, the
and fidelity of the continuous‐variable quantum teleportation security of these classical secret sharing schemes is severely
system at different wavelengths. The numerical simulations compromised [13]. Accordingly, secret sharing has been
indicate that the KNN algorithm has a significant error in the extended into various quantum versions. The first QSS scheme
prediction, while the DT algorithm has a reasonable agreement was proposed by Hillery et al. using the Greenberger–Horne–
with the actual results when the parameters are adjusted Zeilinger (GHZ) state, in which one party can share credible
accordingly. As a result of this approach, authors overcome the classical information with other two or three parties in the
challenges posed by numerical analysis approaches, which presence of an eavesdropper [25]. This was followed by Cleve
show that predictions can be made within an acceptable error et al. proposal of a threshold QSS scheme using the quantum
range under certain conditions. Additionally, their procedure error correction code theory [260]. While various schemes are
addresses quantum channel losses and guides ML‐based constantly being proposed, typically, QSS protocols are divided
continuous‐variable quantum teleportation in practice. Thus, into three categories: when the shared information is an un-
integrating AI and ML algorithms with quantum teleportation known quantum state, it is called quantum state sharing [25,
will shape quantum information processing, especially quan- 234], QSS when shares the classical information [272] and a
tum teleportation as shown in Table 2. Overall, these advances QSS of both of them [273]. These schemes are based on
will lead to the development of a global quantum networked quantum physical properties to share classical information,
landscape in the future. while some schemes are based on quantum mechanics

TABLE 2 Summary of AI/ML applications in quantum entanglement or teleportation schemes.

Quantum protocol or
resource AI/ML model Main objective or outcome
Quantum entanglement Supervised learning Demonstrated a reliable separability‐entanglement classifier [241]

Quantum entanglement Complex‐valued neural network Detected a quantum entanglement in high‐dimensional and multipartite quantum
systems and achieves a detection accuracy of 97.5% [64]

Quantum entanglement Supervised learning Constructed a classifier by employing the relationship between coherence and
entanglement [242]

Quantum teleportation Reinforcement learning þ projective Discovered various quantum communication protocols, including quantum
simulation teleportation, quantum distillation, and end‐to‐end bipartite entanglement
distribution along a chain of quantum repeaters [55]

Quantum teleportation Unsupervised learning Trained an on‐chip autoencoder to encode 3‐D quantum state onto 2‐D quantum
teleportation, observing 3‐D teleportation [62]

Quantum teleportation KNN Predicted the continuous‐variable quantum teleportation squeezing parameter and
the satellite altitude [257]

Abbreviations: AI, artificial intelligence; KNN, K‐nearest neighbour; ML, machine learning.
 26328925, 2024, 3, Downloaded from [Link] by Pondicherry University, Wiley Online Library on [10/08/2025]. See the Terms and Conditions ([Link] on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
216
- MAFU

principles to share the arbitrary quantum state of information j, k, l = 1. At this moment Alice, Bob and Charly each know
[272]. These schemes have been followed by several proposals two of the values i, j, k. When Bob and Charly come together
based on various quantum principles, such as single photons and combine their knowledge, they can establish three of the
[274, 275], entangled states [276] and product states [277]. four parameters. Additionally, they are able to determine the
These schemes include general QSS [278], multiparty QSS [25, last one, which is also known to Alice. By identifying ‘−1’ with
273, 279], multiparty semi‐QSS [280], multiparty to multiparty the bit value ‘0’ and ‘þ1’ with ‘1’, the correlated sequences of
QSS [281, 282], threshold QSS [283, 284], access structure‐ parameter values can then be turned into a secret key.
based QSS [278, 285], circular QSS [286], and dynamic QSS While quantum entanglement constitutes a significant
[287, 288] protocols. resource in QSS protocols, Guo et al. surprisingly demon-
As shown in Figure 9, the QSS scheme uses the GHZ strated the implementation of a QSS protocol without entan-
three‐qubit states [289]. The 3‐qubit entangled GHZ state can glement, in which only product states are employed [290]. This
be described as follows: protocol is more applicable when the number of parties
sharing secrets is large. Notably, the authors demonstrate that
1 this scheme has a theoretic efficiency that can be doubled to
jψ〉GHZ ¼ pffiffi ðj0〉1 j0〉2 j0〉3 þ j1〉1 j1〉2 j1〉3 Þ; ð3Þ
2 approach 100%. Due to the guarantee of quantum funda-
mental principles, QSS offers significantly enhanced security
where |0〉 and |1〉 are orthogonal states in an arbitrary Hilbert than the computationally complex classical secrets. While most
space. Each particle is labelled by one of the indices. As of the proposed QSS schemes are designed based on two‐
illustrated in Figure 9, each of the three photons in the GHZ dimensional quantum systems (i.e. qubits) with multi‐qubit
state has its own mode. When the two beamsplitters located at maximally entangled quantum state, QSS protocol design has
Alice's, Bob's and Charly's have their modes combined, the recently been extended to higher‐dimensional systems (i.e.
probability of finding the three photons in the output ports qudits) [291–293] and orbital angular momentum [294]. This
depends on the settings α, β, γ of the phase shifters: extension allows a higher degree of resource capacities and
enhanced security, leading to improved practical implementa-
1 tions. QSS schemes have an important role to play in quantum
P i;j;k ¼ ð1 þ ijk cosðα þ β þ γÞÞ; ð4Þ
8 communication, such as fibre network configuration [295] and
quantum secure transport system [296]. Specifically, they have
where i, j, k = �1 denotes different output ports. Before each been utilised in a wide range of quantum cryptographic ap-
measurement, Alice, Bob, and Charly will select randomly one plications such as joint checking accounts containing quantum
phase value from (0, π/2). Once a sufficient number of runs money [297], joint financial transaction and quantum digital
have been completed, they identify cases where they all signatures [298], share hard‐to‐create ancilla states [239],
detected a photon. Afterwards, Alice, Bob and Charly perform a distributed computation [297], missile launching
announce the phases chosen and identify cases in which the codes [299], and electronic voting [17, 261, 300].
sum adds up to either 0 or π. Based on the probability function While QSS demonstrates promising potential, the schemes
depicted in Equation (4), a result of 1/4 is obtained for these face several challenges that must be addressed to achieve prac-
cases. Using l = cos(α þ β þ γ) = �1 and Pi,j,k = 1/4 yields i, tical implementation and widespread adoption. As quantum
communication systems are vulnerable to interception, adver-
saries can exploit the vulnerability to intercept and gain unau-
thorised access to shared secrets [275, 277]. These schemes
require precise manipulation and measurement of quantum
states, which requires highly controlled and stable experimental
setups [291, 294]. The shared secrets can be compromised by
external noise, including decoherence and errors in quantum
gates, environmental disturbances, and hardware imperfections.
In addition to technical challenges and communication costs,
practical implementation requires a high level of control and
stability [284]. Furthermore, scalability and efficiency also pose
challenges [274]. What is more, the complexity of the secret‐
sharing scheme typically increases exponentially with the num-
ber of parties involved. As a result, key management, commu-
nication overhead, and computational resources are challenging
for secure sharing and reconstruction of secrets. These risks can
be mitigated with fault‐tolerant quantum computing architec-
tures and robust error correction techniques. As a result,
F I G U R E 9 An illustration of a QSS scheme using the 3‐qubit GHZ
states. A real implementation would have the source as part of Alice setup, developing efficient algorithms and protocols for large‐scale
not that of a fourth party. Reproduced from Ref. [289]. GHZ, secret sharing is now an ongoing research area. A review of
Greenberger–Horne–Zeilinger; QSS, quantum secret sharing. the integration of various ML has been presented, particularly
 26328925, 2024, 3, Downloaded from [Link] by Pondicherry University, Wiley Online Library on [10/08/2025]. See the Terms and Conditions ([Link] on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
MAFU
- 217

concerning multi‐party computation protocols [301]. The au-

thors demonstrate how these algorithms and secret sharing can
be used to safeguard data privacy when multi‐party computation
is being performed.
ML models are typically trained on massive amounts of
data collected from multiple sources [92, 93]. These data
sources may contain sensitive data relating to finance, health-
care, personally identifiable information, and other confidential
data. However, these models are not equipped with privacy‐
preserving features [302]. Rather, the models anonymise the
data, which does not guarantee privacy [303]. This raises
concerns since models must not expose sensitive information.
For these reasons, various ML models incorporate secure MPC
[304], homomorphic encryption [305], and differential privacy
[306] as a means of protecting the privacy of collected data.
Although this enhances security, performance, and accuracy
still needs to be improved. Therefore, we are interested in an
ML model that can achieve both full data privacy for security
and efficiency for practicability. As a result, distributed ML
approaches have been proposed, which allow clients to store
their data locally but train collaboratively. This approach pro-
vides higher accuracy and improved performance, but security
still needs to be improved. As a result, in 2020, Dong et al.
[307] proposed a novel secure computation framework that F I G U R E 1 0 An illustration of the proposed efficient three‐party
integrates distributed ML, specifically linear regression, multi- computation protocol for secret sharing‐based K‐means clustering
layer perceptron, and CNN with secret sharing to achieve algorithm. Based on the definition of full data privacy, the information
about the intermediate centroids, cluster assignments, and cluster sizes
reliable performance, accuracy and high‐level security in semi‐
should be protected. Thus, the only information that can be leaked is the
dishonest parties. This work demonstrates an efficient practical range of Dk. Hence, the protocol provides full data privacy. Reproduced
system that can be used to jointly learn an accurate model from Ref. [308].
under semi‐honest and server‐only malicious adversary secu-
rity, respectively. Notably, a client may not learn more infor-
TABLE 3 A description of datasets used in the experiments.
mation than what is contained in the trained model, and a
parameter server may not learn sensitive information. As a Dataset n K d Accuracy
result, integrating distributed ML with secret‐sharing schemes Iris 150 3 4 92.67%
ensures that the model achieves the best overall performance
Arff 100 4 2 98.20%
while still meeting security requirements.
Furthermore, in 2022, Wei et al. [308] presented a privacy‐ Self‐generated {10, 000, 100, 000} {2, 5} {5, 10, 15, 20} ‐
preserving K‐means clustering algorithm that employs repli- Note: Each dataset has n data points, K clusters, and d dimensions. The accuracy of
cated secret sharing [309] with the honest majority in semi‐ different datasets is also evaluated if a ground truth model exists for the dataset.
honest models. The secret sharing‐based privacy‐preserving Reproduced from Ref. [308].

K‐means clustering scheme is shown in Figure 10. Particu-

larly, a privacy‐preserving K‐means clustering, which has full
data privacy, allows the parties to cluster their combined arff and 4D dataset Iris to evaluate accuracy. Particularly, they
datasets without revealing any other information except for the compare the accuracy of ground truth, plaintext and privacy‐
final centroid [310]. This means the information on interme- preserving models for 2D dataset arff and obtain a value of
diate centroids, cluster assignments, and cluster sizes should be 98.20%. The comparison of the experimental results of this
protected in the protocol. Therefore, the clustering task is dataset is visually illustrated in Figure 11. However, for the iris
outsourced to three semi‐honest computing servers. Specif- dataset, the privacy‐preserving model reaches an accuracy of
ically, the proposed protocol provides full privacy guarantees, 92.67%. Therefore, the proposed scheme achieves both full
which allows different computing parties to cluster the com- data privacy for security and efficiency for practicability.
bined datasets without revealing any other information except Experimental results demonstrate that the proposed pro-
the final centroids. As a result, the protocol is secure against a tocol achieves the same level of accuracy as plaintext K‐means
single corrupt server under a semi‐honest model. The pro- clustering. By using the fast network, the privacy‐preserving
posed secret sharing‐based privacy‐preserving K‐means clus- scheme can deal with datasets of millions of points in a
tering scheme is illustrated in Figure 10. The scheme is reasonable amount of time. Comparatively, this scheme ach-
implemented and analysed and the experimental results are ieves about 16.5–25.2 times higher computation speeds and
shown in Table 3. Particularly, the authors use the 2D dataset 63.8–68.0 times lower communication rates than the existing
 26328925, 2024, 3, Downloaded from [Link] by Pondicherry University, Wiley Online Library on [10/08/2025]. See the Terms and Conditions ([Link] on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
218
- MAFU

privacy‐preserving schemes. Additionally to its high efficiency result, to enable long‐distance end‐to‐end communication of
and practicality, the scheme is also suitable for large‐scale qubits, quantum networks rely on the following three primary
clustering tasks. Overall, the privacy‐preserving scheme gua- components: end nodes, which serve as processors, quantum
rantees full data privacy and is suitable for secure outsourcing channels, which distribute information, and quantum repeaters,
of computations based on secret‐sharing. The ability of ML to which generate separate entanglement pairs and transmit
optimise resource allocation, enhance security, identify vul- qubits from each pair to the respective nodes [20, 21]. Thus, in
nerabilities, mitigate noise and imperfections, and uncover this section, we will discuss the integration of AI and ML al-
hidden correlations makes it a valuable tool for increasing the gorithms to address the aforementioned challenges or limita-
effectiveness and reliability of various QSS schemes. Even tions and improve the performance of these components,
though this field is still undergoing development as summar- resulting in an overall improvement in quantum networks. An
ised in Table 4, these few examples demonstrate the potential illustration of a typical quantum network is presented in
of integrating ML into secret‐sharing schemes show the po- Figure 12. Quantum network systems rely on existing network
tential for improving QSS. Therefore, further research into infrastructure for exchanging classical messages in order to run
incorporating AI and ML into QSS may ultimately result in quantum protocols, as well as for controlling and managing the
more robust and secure methods for sharing secrets in the network [311]. Due to challenges in building long‐distance
quantum world. communication, a chain of automated quantum repeaters is
used in order to build links due to the limitations associated
with sending secure messages over long distances. Regarding a
3.4 | Quantum networks comprehensive review of quantum networks, we refer the
reader to Refs [21, 312–315].
A quantum network is intended to facilitate secure commu- Particularly, quantum repeaters are hybrid devices that can
nication between quantum computers or secure cloud quantum broadcast ‘flying’ qubits (photons) without measuring or
computing, as well as quantum‐enhanced measurement net- cloning them, which is against the quantum no‐cloning theo-
works, ultimately leading to a quantum internet, a global rem. The operation of quantum repeaters is intrinsically based
network capable of transmitting quantum data [20, 311]. Thus, on the phenomenon of quantum teleportation, one of the
quantum networks promise to enhance existing classical net- most fascinating applications of entanglement. Typically, a
works as well as execute protocols that would be impossible in quantum repeater comprises three elements: entanglement
a classical network. While significant progress has been made, generation, entanglement connection, and entanglement puri-
challenges remain, including transmission losses, decoherence, fication. By connecting repeaters, entanglement can be
and limitations imposed by the no‐cloning theorem. As a

F I G U R E 1 1 The comparison of accuracy for ground truth, plaintext

and privacy‐preserving for secret sharing‐based K‐means schemes using the F I G U R E 1 2 An illustration of a typical quantum network. The system
2D dataset arff. (a) a ground truth model and (b) a plaintext and privacy‐ consists of several components including a quantum repeater, classical
preserving model. Notably, the privacy‐preserving model achieves the same links, and quantum links for establishing long‐distance quantum
accuracy level as a plaintext model. The accuracy is 98.20% compared to communication or large‐scale quantum networks. Reproduced from
the ground truth model. Reproduced from Ref. [308]. Ref. [311].

TABLE 4 Summary of AI/ML applications in secret sharing schemes.

Scheme AI/ML model Main objective or outcome

Secret sharing Linear regression, MLP, and CNN Improved the accuracy, reliability, security, and performance of privacy‐preserving distributed
ML based on secret sharing [307]

Secret sharing K‐means clustering Enhanced the efficiency and practicality of a privacy‐preserving secret‐sharing‐based secure
three‐party computation [308]

Abbreviations: AI, artificial intelligence; CNN, convolutional neural network; ML, machine learning; MLP, multilayer perceptron.
 26328925, 2024, 3, Downloaded from [Link] by Pondicherry University, Wiley Online Library on [10/08/2025]. See the Terms and Conditions ([Link] on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
MAFU
- 219

extended by a significant distance, providing a viable option for reason for this is that optical photons have low quantum in-
establishing long‐distance quantum communication or large‐ formation transmission losses and negligible quantum deco-
scale quantum networks [214, 222]. herence in space. This represents a significant development
An example of a long‐distance end‐end‐QKD based on that shows the possibility of building quantum networks across
trusted quantum repeaters is shown in Figure 13. Typically, the continents.
sender and receiver are connected via a quantum channel and a Due to quantum properties such as quantum entangle-
public channel. The two intermediate nodes with trusted re- ment, quantum uncertainty, and quantum no‐cloning theorem,
peaters are placed between the source node and the destination quantum networks can provide ultra‐secure information
node. Figure 13 illustrates how three secret keys are produced exchanged over multiple technologies across numerous chan-
between the source node and intermediate node 1, interme- nels [20, 21, 247]. With advances in quantum computing
diate node 1 and intermediate node 2, intermediate node 2, and technologies, quantum networks can now complement classical
the destination node. Most significantly, the produced three networks in processing and delivering information securely
secret keys have the same key size. Essentially, the procedure since not all information transmitted must be encoded in
consists of three steps: (1) Intermediate node 1 encrypts Ks1 photons, for example, quantum teleportation [23, 31, 53].
using K12 and obtains the ciphertext Ks1 ⊕ K12. Then, the Thus, combining classical and quantum networks to create a
ciphertext Ks1 ⊕ K12 is sent to intermediate node 2. (2) Using quantum internet will result in capabilities that are unparallelled
K12, intermediate node 2 decrypts Ks1 ⊕ K12 to obtain Ks1. by either technology alone [20]. For instance, the quantum
Then, intermediate node 2 encrypts Ks1 using K2d, to obtain internet will provide new encryption services, enhance sensor
the ciphertext Ks1 ⊗ K2d. The ciphertext Ks1 ⊗ K2d is sent to network sensitivity, and connect distant quantum computers to
the destination node. (3) Using K2d, the destination node de- facilitate computation, share quantum data, and expand the
crypts the ciphertext Ks1 ⊗ K2d to obtain Ks1. As a result of the class of complex problems that can be solved [311, 325].
above three steps, Ks1 can be shared between the source node Due to photon limitations, such as losses at more
and the destination node. Notably, for end‐to‐end QKD over considerable distances, quantum networks contain vital device
long distances, several intermediate nodes with trusted re- elements, including quantum nodes [326], quantum repeaters
peaters can be placed between the source and destination or quantum switches [155, 156], quantum memories, and
nodes, where all of the intermediate nodes must be trustworthy quantum channels [327, 328]. The purpose of these devices is
because they will all know the real secret keys, for example, Ks1. to unlock the full potential of quantum networks. As a result,
Although using trusted repeaters is a notable advance, despite their potential, unfortunately, some of these devices are
there are still challenges in transmitting information over large not capable of being successfully deployed with current tech-
distances due to turbulence and losses in the physical channels nology. This presents an opportunity for AI and ML to fill the
[17, 53]. The advancements in integrated photonics remain gap. Quantum networks have the following applications:
among the most stable, compact, and robust platforms for quantum‐secure communications [2, 3, 16, 126, 139], secure
miniaturising massive photonic circuits [317, 318]. Due to the identification [329], blind quantum computing [330, 331],
compatibility of integrated photonic chips with current inte- network clock synchronisation [332], distributed quantum
grated photonic telecommunication hardware, seamless inte- computing [333, 334], and entanglement sensor network [335].
gration is assured with classical communication channels and Notably, despite challenges and limitations, significant progress
transceivers [319]. Through this integration, developments are has been made in developing quantum networks for sending
underway to realise hybrid classical and quantum communi- quantum information over reasonable distances, and they are
cations, with a current distance of more than 100 km achieved now available for use in QKD with trusted nodes [21].
[320–322]. Notably, an integrated satellite‐ground quantum While quantum networks hold significant potential for
communication network of over 4600 km has recently been secure communication and quantum computing, the develop-
reported [323]. The capability of satellite‐ground QKD using ment and deployment of quantum information in a large‐scale
optical photons shows tremendous potential for long‐distance network face several significant challenges. These challenges
and intercontinental quantum communication [324]. The include: maintaining quantum entanglement, maintaining
quantum coherence, scaling and interoperability, distance and
connectivity, security, and integration with classical communi-
cation infrastructure [21, 227, 311]. Moreover, there are con-
cerns regarding the nature of the hardware platform and its
precise implementation requirements. As part of these chal-
lenges, scientists and researchers are actively addressing the
issue of storing and manipulating qubits by integrating quan-
tum error correction, quantum repeaters, quantum memories,
and quantum routers into quantum networks [53, 156, 313,
326, 336]. Among these efforts is a recent attempt to deter-
F I G U R E 1 3 An illustration of a typical trusted quantum repeater. An
example of QKP for managing secret keys in a pair‐wise manner between mine whether a quantum network of nodes connected by
QKD node 1 and QKD node 2. Reproduced from Ref. [316]. QKD, quantum links has reached a particular development stage. This
quantum key distribution; QKP, quantum key pool. is done by evaluating the robustness of noisy quantum
 26328925, 2024, 3, Downloaded from [Link] by Pondicherry University, Wiley Online Library on [10/08/2025]. See the Terms and Conditions ([Link] on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
220
- MAFU

networks [337] and quantum network benchmarking. By using network infrastructure. Considering that tenant requests arrive
this process, it is possible to benchmark the quality of a on a dynamic basis, obtaining secret keys from the QKD
quantum network and measure its fidelity [338, 339]. Due to network infrastructure becomes a challenge when simulta-
the inherent complexity of manipulating quantum systems and neously adapting to the complicated network conditions of the
their vulnerability to noise, some challenges remain [340]. For QKD network. As a result, there is an urgent need to develop
instance, a quantum repeater requires a quantum memory, a an effective online multi‐tenant secret‐key assignment (MTKA)
device that stores conjugate quantum variables. However, no process that can accommodate multiple tenant requests in
reliable and practical quantum memory is currently available QKD networks. According to Cao et al. [353], reinforcement
[53]. As an intermediate step, trusted repeaters over long dis- learning algorithms can efficiently allocate several network
tances [166, 340–342], MDI‐QKD networks [53], and TF‐ resources, including secret keys, to achieve a MTKA algorithm
QKD protocols [138] are being implemented, and distances for QKD networks. According to the results, reinforcement
of up to 830‐km reported [343]. These advances constitute learning based on the MTKA algorithm significantly out-
significant steps towards the development of the quantum performs benchmark heuristics (e.g. random, fit, best fit)
internet. regarding tenant‐request blocking probability and secret‐key
Scalable quantum networks require various optimisation resource utilisation. The reinforcement learning algorithm
models for efficiently assigning quantum information and re- based on the MTKA algorithm effectively reduces the likeli-
peaters in quantum networks. Various efforts have been made hood of tenant requests being blocked by more than half when
to integrate AI and ML into quantum networks leading to compared to the benchmark heuristics. Furthermore, a recent
improved noise mitigation, network self‐configuration, study by Sharma et al. [120] illustrates how reinforcement
resource allocation efficiency, reliable error correction, self‐ learning algorithms can be utilised to address the routing and
optimisation, increased performance and security [129, 344– resource assignment challenge in quantum signal channels for
346]. This has resulted in increased performance and secu- QKD optical networks. The authors present a Deep rein-
rity, particularly concerning QKD‐secured optical networks. forcement learning (DRL)‐based routing and resource assign-
For instance, Ou et al. [189] demonstrated how ML‐aided ment approach that relies upon proximal policy optimisation in
software‐defined networking relying on optimal resource which the optimal route is selected and network resources are
allocation was constructed for investigating the coexistence of efficiently utilised to satisfy quantum signal channel resource
hybrid quantum and classical channels in a QKD‐integrated requirements for QKD lightpath requests in QKD optical
optical network. The authors present various supervised ML networks. A major finding of the study is that the DRL‐based
methods to estimate channel QKD performance (noise, SKR, routine and resource assignment scheme outperforms the
and quantum bit error rate) when Ch‐Cs in various quantities, deep‐Q network method, as well as first‐fit and random‐fit
spectrum allocations, launch power, and channel spacing are baseline schemes in terms of both blocking probability and
present in the C‐band. As compared to various models such as resource utilisation.
RF, least‐squares method, lasso regression, and ridge regres- Quantum memories constitute an integral component of
sion, the KNN was the most accurate model. Most signifi- quantum repeaters, which enables long‐distance quantum
cantly, this work demonstrates the possibility of QKD communication and photonic quantum computation protocols
integration with conventional DWDM fibre networks, which is beyond that permissible by passive transmission [354]. A uni-
crucial for the commercialisation and implementation of versal optical quantum memory can store unknown input
quantum networks. states of light and release them on demand with high efficiency
AI and ML techniques have been proposed for predicting and without additional noise [355, 356]. A critical threshold of
optimal channel allocation and quantum parameter values in 50% total storage and recall efficiency may be required
near real‐time, enabling reallocation of quantum channels while depending on the application; however, coherence time will be
ensuring excellent performance through the efficient evalua- restricted to tens of microseconds [356]. Over the years,
tion of parameter values [347–350]. Furthermore, various low‐ various mechanisms for implementing optical quantum mem-
complexity, but near‐optimal wavelength assignment methods ory have been explored and developed. These include elec-
have been employed to increase the achievable secret key rate tronic induced transparency [357], Raman schemes [358], as
in hybrid quantum‐classical networks [228, 351, 352]. Partic- well as atomic frequency combs [359]. However, among other
ularly, due to their non‐reusable nature, secret key resources in schemes, the gradient echo memory has demonstrated the
QKD networks are essentially finite and valuable, making their greatest recall frequency with longer coherence times [356,
assignment unique. Thus, the expansive cost and challenges 360]. Several ML methods have been applied, resulting in
inherent in deploying QKD networks, multi‐tenancy has significant improvements in optical depth. Accordingly, to
become an attractive solution to improve cost efficiency for enhance these schemes, Leung et al. [361], the integration of
future QKD networks, particularly for organisations with high‐ ML and compatible single photons improves gradient echo
security demands [316]. Essentially, a multi‐tenant QKD memory performance and may facilitate quantum computation
network is composed of multiple QKD tenants who can share by using it as quantum gates. Due to the analytical complexity
the same QKD network infrastructure in order to obtain their of the system, it is generally difficult to optimise a cold atomic
keys for securing their data transfers. Specifically, the generated ensemble to achieve increased optical depth and lower tem-
secret keys can be allocated to multiple tenants over a QKD peratures. A large optical depth should result in higher memory
 26328925, 2024, 3, Downloaded from [Link] by Pondicherry University, Wiley Online Library on [10/08/2025]. See the Terms and Conditions ([Link] on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
MAFU
- 221

efficiencies with stronger atom‐light interactions. Particularly, known as memory cutoff. This technique allows one to explore
the authors employ reinforcement learning to optimise atom general cut‐off strategies that dynamically adjust to quantum
trapping and develop high‐extinction filtering methods to repeater states. A recent paper by Robertson et al. [365] dem-
isolate single photons from bright pump light. As a result, the onstrates how a genetic algorithm may be used to optimise the
cold atomic ensemble is optimised for higher optical depth and write control for Gaussian signal pulses of an optical memory.
lower temperature, which improves the efficiency and coher- This work shows that it is possible to reduce pulse energy by 30%
ence times of gradient echo memories. without compromising efficiency. The results of this study are
Moreover, Buchler et al. [362] use an ANN to optimise a consistent with those of Shinbrough et al. in Ref. [366] and
laser‐cooled atomic ensemble of rubidium 87 atoms to achieve support the development of optical quantum memories with
the ‘stopped and stationary light’ optical quantum memory high efficiency. A significant aspect of quantum networks is their
with an efficiency of about 87% with a decay time of 1 ms with evolution towards practical maturity, which makes stand-
a cold‐atom memory. An ML algorithm is used to expedite the ardisation efforts increasingly crucial. Among these efforts are
trial‐and‐error approach when maximising the on‐axis optical the standardisation of AI and ML‐enabled QKD networks [129,
depth of the magneto–optic trap. Despite various efforts to 367]. The standardisation efforts are intended to address the
develop efficient quantum networks, quantum memories still requirements, architectures, functional capabilities, and appli-
have limited coherence times. Although various efforts have cation programming interfaces needed to implement converged
been made to develop efficient quantum networks, quantum future networks. Accordingly, Table 5 summarises the progress
memories still have a limited coherence time. Khatri suggests in applying AI/ML algorithms in quantum network schemes.
that this challenge can be addressed by developing optimal
entanglement distribution protocols that take into account the
limitations of current and near‐term hardware [363]. This in- 4 | DISCUSSION AND CONCLUSION
cludes quantum memories with limited coherence times. The
author investigates quantum network protocols for entangle- Quantum communication is arguably one of the most suc-
ment distribution using decision processes theory. To find the cessful fields that have evolved from quantum theory. Ad-
optimal protocol the author leverages dynamic programming vances in research and development have resulted in the
or reinforcement learning algorithms. Notably, the author commercialisation of several quantum communication tech-
optimises entanglement times on quantum channels across nologies that have transformed a variety of industries. Addi-
paths in order to establish entanglement states between two tionally, standards and certification processes have been
end nodes prior to the decay of the channel. The findings of developed to address regulatory considerations, interopera-
this research pave the way for a systematic investigation of the bility, performance, and security requirements. This facilitates
limitations of near‐term quantum networks as well as their the integration of quantum systems into existing industries. A
physical realisation requirements. certification framework fosters the adoption and acceptance of
Yun‐Hong et al. developed a method to predict the key rate quantum communication products and services by ensuring
for satellite‐to‐ground quantum communication channels via quality and reliability. Quantum communication relies on
ML and stellar image recognition in quantum networks [364]. manipulating quantum states in order to ensure reliable and
This work demonstrates a stellar image recognition accuracy rate efficient transmission of quantum information. Although sig-
of approximately 88% and recommends whether to accept or nificant progress has been made in this area, a number of
abort the protocol based on the predicted key rate. Specifically, if fundamental challenges still need to be addressed in order to
satellite‐to‐ground QKD is accepted, the average rate of sifted accomplish this goal. Towards addressing these challenges, this
key at an elevation angle of 39.5° is 8.8 kbit/s, which is adequate work has examined several key milestones demonstrating how
for satellite‐to‐ground QKD tasks involving multiple ground AI and ML algorithms have been integrated to optimise
stations. Considering the finite nature of QKD resources, the various components or implementation parameters to enhance
present work improves the prediction and experimental verifi- quantum communication protocols and quantum networks,
cation of satellite‐to‐ground quantum communication networks. leading to large‐scale deployment and ultimately shaping both
A deep NN technique was presented in 2022 by Le et al. that took fields. Moreover, this has the potential to open up new avenues
account of the current network state to schedule the network's for improving customer experience, optimising network op-
demands, which were then routed by a qubit‐preserved short‐ erations, and increasing the efficiency of a network.
path algorithm [328]. One of the critical challenges of next‐ The discussion on QKD demonstrates examples of how
generation network systems is routing in quantum networks. AI and ML algorithms have been utilised to predict physical
The results indicate that in a qubit‐limited grid network, the deep parameters, improve the accuracy of raw key data, predict
reinforcement routing scheme maintains an average fulfiling rate quantum channel quality or failure due to a disturbance as well
of 80% and approximately 60% of routed requests under as recover messages and images with a high level of accuracy.
extreme conditions. According to the authors, deep reinforce- These intelligent techniques adjust communication parameters
ment routing schemes have polynomial complexity and in real‐time and improve the overall key generation and man-
computational time based on quantum network sizes. Recently, agement, security, and efficiency of the protocol. Furthermore,
Reiß et al. [69] illustrates how DRL can be used to search for AI and ML methods have been used to detect and mitigate
optimal solutions for the memory storage time limits, typically various types of attacks, thereby ensuring the integrity of
 26328925, 2024, 3, Downloaded from [Link] by Pondicherry University, Wiley Online Library on [10/08/2025]. See the Terms and Conditions ([Link] on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
222
- MAFU

TABLE 5 Summary of AI/ML applications in quantum network schemes.

Quantum network or device AI/ML model Main objective or outcome

QKD þ DWDM fibre network KNN To investigate and optimise the coexistence of hybrid quantum‐classical channels in a QKD‐
integrated optical network [189]

Hybrid quantum‐classical networks Reinforcement learning To increase the number of achievable secret keys for securing data transfers in a multi‐tenant
QKD network architecture [316, 353]

Optical networks DRL Optimise routing and resource assignment problem in the quantum signal channel of QKD‐
secured optical network [120]

Quantum memory ANN To optimise a laser‐cooled atomic ensemble to achieve ‘stopped and stationary light’ optical
quantum memory with an efficiency of about 87% [362]

Quantum memory Reinforcement learning To develop the mathematical tools needed to perform reinforcement learning of entanglement
distribution protocols in general quantum networks [363]

Quantum network ML To accurately and quickly predict the key rate for satellite‐to‐ground quantum communication
channels via ML and stellar image recognition in quantum network [220]

Quantum repeaters DRL To search for optimal solutions to memory storage time limits [69]

Abbreviations: AI, artificial intelligence; ANN, artificial neural network; DRL, deep reinforcement learning; DWDM, dense wavelength‐division multiplexing; KNN, K‐nearest
neighbour; ML, machine learning; QKD, quantum key distribution.

quantum communications. Most significantly, this has also Considering the challenges in achieving long‐distance
resulted in enhanced security, improved efficiency, and ensured transmission, we have explored how AI and ML have been
the confidentiality and integrity of information. applied in quantum communication networks. AI and ML al-
We discuss implementations demonstrating the use of AI gorithms have been used to optimise the existence of hybrid
and ML to mitigate noise and decoherence, leading to quantum‐classical channels, optimise routing and resource
enhanced performance of quantum teleportation schemes. allocation, optimise solutions to memory storage time limits,
Various works have shown that AI and ML techniques can and improve scalability. As quantum communication networks
improve several quantum error correction codes, making evolve, AI and ML will be crucial to unlocking their potential.
quantum teleportation systems more fault‐tolerant and reliable. Besides this significant progress, some limitations must be
In particular, AI and ML algorithms have been used to identify overcome in order to harness the full potential of integrating
central quantum protocols as well as detect and classify AI and ML in quantum communication. For instance, a key
quantum entanglement. This could reduce the risk of eaves- aspect of AI and ML is the need for vast amounts of data to
dropping and maintain confidentiality while enabling secure train models. However, the quantum communication field
data transmission. Leveraging various AI and ML techniques often deals with sensitive data, such as encryption keys, that
leads to improved efficiency, reliability, and security of quan- require stringent security measures. Balancing the need for data
tum teleportation protocols. This could enable longer and with the desire for privacy poses a significant challenge.
more secure quantum communication. Furthermore, quantum communication schemes, including
Several authors discuss how ML algorithms can be used to quantum networks are inherently susceptible to noise, dis-
enhance data transmission rates and optimise the allocation of rupting the transmission of information. Developing algo-
network resources based on the specific resource requirements rithms that can efficiently handle and mitigate quantum noise is
of different communication tasks. The results demonstrate a crucial for the successful integration of AI and ML in quantum
significant improvement in resource utilisation and a reduction communication. Moreover, integrating AI and ML algorithms
in the overall cost of quantum communication systems. Notably, with quantum communication requires optimising them for
noise mitigation is imperative for all quantum communication quantum hardware. This optimisation process can be chal-
protocols and networks since they affect scalability. It has been lenging due to the limited computational power currently
demonstrated that ML algorithms can be trained to adjust QSS available in quantum computers.
parameters dynamically based on real‐time feedback and miti- As a future direction, it is critical to continue to expand
gate noise. This has led to improved performance, reliability, efforts to improve the performance of quantum devices
security, and defence capabilities of QSS schemes against while overcoming their limitations. Moreover, improving the
emerging threats. AI and ML algorithms can predict and predictability of quantum systems will help AI and ML al-
compensate for signal degradation or noise by analysing quan- gorithms make more accurate predictions and perform
tum channels. This will enhance the efficiency and practicality of efficient quantum computations. Another key aspect is
various privacy‐preserving secret‐sharing‐based schemes. To- developing algorithms to efficiently process and interpret
wards the future, AI and ML have a significant potential to quantum data. These algorithms need to be robust, scalable,
optimise various quantum communication channels used in se- and capable of handling the complex nature of quantum
cret sharing protocols. information. While there are still challenges and limitations
 26328925, 2024, 3, Downloaded from [Link] by Pondicherry University, Wiley Online Library on [10/08/2025]. See the Terms and Conditions ([Link] on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
MAFU
- 223

in quantum communication, with continued advancements in 14. Metcalf, B.J., et al.: Quantum teleportation on a photonic chip. Nat.
AI and ML, this field is poised for significant progress in Photonics 8(10), 770–774 (2014). [Link]
the future. This review sets the stage for delivering the 2014.217
15. Luo, W., et al.: Recent progress in quantum photonic chips for quantum
envisaged secure quantum internet. communication and internet. Light Sci. Appl. 12(1), 175 (2023). https://
[Link]/10.1038/s41377‐023‐01173‐8
AUT HO R CO NT R I BU T I ON S 16. Bennett, C., Brassard, G.I.: In: Proceedings of the IEEE International
The author confirms being the sole contributor to this work Conference on Computers Systems and Signal Processing Bangalore
and has approved it for publication. India, pp. 175 (1984)
17. Pirandola, S., et al.: Advances in quantum cryptography. Adv. Opt.
Photon. 12(4), 1012–1236 (2020). [Link]
CON FLIC T OF I N T ER E ST STAT E M EN T 18. Ladd, T.D., et al.: Quantum computers. Nature 464(7285), 45–53 (2010).
The author declares that they have no competing interests. [Link]
19. Pompili, M., et al.: Realization of a multinode quantum network of
remote solid‐state qubits. Science 372(6539), 259–264 (2021). https://
DATA AVA IL AB I LI T Y S TAT E M EN T
[Link]/10.1126/science.abg1919
The original contributions presented in the study are included 20. Kimble, H.J.: The quantum internet. Nature 453(7198), 1023–1030
in the article. (2008). [Link]
21. Wehner, S., Elkouss, D., Hanson, R.: Quantum internet: a vision for the
CON SE N T TO PART I CI PA T E road ahead. Science 362(6412), eaam9288 (2018). [Link]
Not applicable. 1126/science.aam9288
22. Sidhu, J.S., et al.: Advances in space quantum communications. IET
Quantum Commun. 2(4), 182–217 (2021). [Link]
CON SE NT FO R PU BL I CAT I ON qtc2.12015
Not applicable. 23. Pirandola, S., Braunstein, S.L.: Physics: unite to build a quantum
internet. Nature 532(7598), 169–171 (2016). [Link]
O RC ID 532169a
24. Bennett, C.H., et al.: Teleporting an unknown quantum state via dual
Mhlambululi Mafu [Link] classical and Einstein‐Podolsky‐Rosen channels. Phys. Rev. Lett. 70(13),
1895–1899 (1993). [Link]
R EF ER E NCE S 25. Hillery, M., Bužek, V., Berthiaume, A.: Quantum secret sharing. Phys. Rev.
1. DiVincenzo, D.P.: The physical implementation of quantum compu- 59(3), 1829–1834 (1999). [Link]
tation. Fortschr. Phys. Prog. Phys. 48(9–11), 771–783 (2000). https:// 26. Mohseni, M., et al.: Commercialize quantum technologies in five years.
[Link]/10.1002/1521‐3978(200009)48:9/11<771::aid‐prop771>3.0. Nature 543(7644), 171–174 (2017). [Link]
co;2‐e 27. Gibney, E.: The quantum gold rush. Nature 574(7776), 22–24 (2019).
2. Gisin, N., Thew, R.: Quantum communication. Nat. Photonics 1(3), [Link]
165–171 (2007). [Link] 28. Bayerstadler, A., et al.: Industry quantum computing applications. EPJ
3. Gisin, N., et al.: Quantum cryptography. Rev. Mod. Phys. 74(1), Quantum Technol. 8(1), 25 (2021). [Link]
145–195 (2002). [Link] s40507‐021‐00114‐x
4. Nielsen, M.A., Chuang, I.L.: Quantum Computation and Quantum 29. van Deventer, O., et al.: Towards European standards for quantum
Information. Cambridge: Cambridge University Press (2010) technologies. EPJ Quantum Technol. 9(1), 33 (2022). [Link]
5. Horodecki, R., et al.: Quantum entanglement. Rev. Mod. Phys. 81(2), 10.1140/epjqt/s40507‐022‐00150‐1
865–942 (2009). [Link] 30. T. I. O. for Standardization: ISO/IEC 23837‐1:2023 information se-
6. Bouwmeester, D., Zeilinger, A.: The physics of quantum information: curity security requirements, test and evaluation methods for quantum
basic concepts. In: The Physics of Quantum Information: Quantum key distribution (2023)
Cryptography, Quantum Teleportation, Quantum Computation, pp. 31. Singh, A., et al.: Quantum internet—applications, functionalities,
1–14. Springer (2000) enabling technologies, challenges, and research directions. IEEE
7. Busch, P., Heinonen, T., Lahti, P.: Heisenberg’s uncertainty principle. Commun. Surv. Tutorials 23(4), 2218–2247 (2021). [Link]
Phys. Rep. 452(6), 155–176 (2007). [Link] 1109/comst.2021.3109944
2007.05.006 32. Ramakrishnan, R.K., et al.: The quantum internet: a hardware review. J.
8. Wootters, W.K., Zurek, W.H.: A single quantum cannot be cloned. Indian Inst. Sci. 103(2), 1–21 (2022). [Link]
Nature 299(5886), 802–803 (1982). [Link] 022‐00336‐7
9. Scarani, V., et al.: The security of practical quantum key distribution. 33. Rohde, P.P.: The Quantum Internet: The Second Quantum Revolution.
Rev. Mod. Phys. 81(3), 1301–1350 (2009). [Link] Cambridge University Press (2021)
revmodphys.81.1301 34. Mafu, M., Senekane, M.: Quantum technology for development
10. Rivest, R.L., Shamir, A., Adleman, L.: A method for obtaining digital framework as a tool for science diplomacy. Front. Res. Metr. Anal. 8
signatures and public‐key cryptosystems. Commun. ACM 21(2), (2023). [Link]
120–126 (1978). [Link] 35. Xu, Y., et al.: Artificial intelligence: a powerful paradigm for scientific
11. Renner, R., Wolf, R.: Quantum advantage in cryptography. AIAA J. research. Innovation 2(4), 100179 (2021). [Link]
61(5), 1895–1910 (2023). [Link] xinn.2021.100179
12. Chen, L., Chen, L., Jordan, S., Liu, Y.‐K., Moody, D., Peralta, R., 36. Shalev‐Shwartz, S., Ben‐David, S.: Understanding Machine Learning:
Perlner, R.A., Smith‐Tone, D.: Report on Post‐Quantum Cryptography, From Theory to Algorithms. Cambridge University Press (2014)
vol. 12. US Department of Commerce, National Institute of Standards 37. Davenport, T.H., Ronanki, R.: Artificial intelligence for the real world.
and Technology … (2016). [Link] Harv. Bus. Rev. 96(1), 108–116 (2018)
13. Shor, P.W.: Algorithms for quantum computation: discrete logarithms 38. Harrington, P.: Machine Learning in Action. Simon and Schuster (2012)
and factoring. In: Proceedings 35th Annual Symposium on Foundations 39. Lloyd, S.: Quantum machine learning for data classification. Physics 14,
of Computer Science, pp. 124–134. IEEE (1994) 79 (2021). [Link]
 26328925, 2024, 3, Downloaded from [Link] by Pondicherry University, Wiley Online Library on [10/08/2025]. See the Terms and Conditions ([Link] on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
224
- MAFU

40. Schuld, M., Petruccione, F.: Machine Learning with Quantum Com- 62. Zhang, H., et al.: Resource‐efficient high‐dimensional subspace tele-
puters. Springer (2021) portation with a quantum autoencoder. Sci. Adv. 8(40), eabn9783
41. Dean, J.: Big Data, Data Mining, and Machine Learning: Value Creation (2022). [Link]
for Business Leaders and Practitioners. John Wiley & Sons (2014) 63. Hu, X.‐M., et al.: Progress in quantum teleportation. Nat. Rev. Phys.
42. Wang, M.: CAS center for excellence in quantum information and 5(6), 1–15 (2023). [Link]
quantum physics: exploring frontiers of quantum physics and quantum 64. Chen, Y., et al.: Detecting quantum entanglement with unsupervised
technology. Natl. Sci. Rev. 4(1), 144–152 (2017). [Link] learning. Quantum Sci. Technol. 7(1), 015005 (2021). [Link]
1093/nsr/nwx025 10.1088/2058‐9565/ac310f
43. Hasan, S.R., et al.: Quantum communication systems: vision, protocols, 65. Liu, Y., et al.: Entanglement‐based feature extraction by tensor network
applications, and challenges. IEEE Access 11, 15855–15877 (2023). machine learning. Front. Appl. Math. Stat. 7, 716044 (2021). https://
[Link] [Link]/10.3389/fams.2021.716044
44. Bin, Y., Changwei, H., Xizheng, Y.: Research on business support sys- 66. Bai, S.‐C., Tang, Y.‐C., Ran, S.‐J.: Unsupervised recognition of infor-
tem of quantum communication. Inf. Commun. Technol. Policy 45(9), mative features via tensor network machine learning and quantum
72 (2019) entanglement variations. Chin. Phys. Lett. 39(10), 100701 (2022).
45. Khan, F.S., La Torre, D.: Quantum information technology and inno- [Link]
vation: a brief history, current state and future perspectives for business 67. Pawłowski, J., Krawczyk, M.: Quantification of entanglement with sia-
and management. Technol. Anal. Strat. Manag. 33(11), 1281–1289 mese convolutional neural networks. arXiv preprint arXiv:2210.07410
(2021). [Link] (2022)
46. Humble, T.: Consumer applications of quantum computing: a prom- 68. Sheng, Y.‐B., Zhou, L.: Distributed secure quantum machine learning.
ising approach for secure computation, trusted data storage, and effi- Sci. Bull. 62(14), 1025–1029 (2017). [Link]
cient applications. IEEE Consum. Electron. Mag. 7(6), 8–14 (2018). 2017.06.007
[Link] 69. Reiß, S.D., van Loock, P.: Deep reinforcement learning for key distri-
47. Sotelo, R.: Quantum in consumer technology. IEEE Consum. Electron. bution based on quantum repeaters. Phys. Rev. 108(1), 012406 (2023).
Mag. 12(5), 4–7 (2023). [Link] [Link]
48. Fox, M.F., Zwickl, B.M., Lewandowski, H.: Preparing for the quantum 70. Marquardt, F.: Machine learning and quantum devices. SciPost Phys.
revolution: what is the role of higher education? Phys. Rev. Phys. Educ. Lecture Notes, 029 (2021). [Link]
Re s. 1 6( 2 ) , 0 2 0 1 3 1 ( 2 0 2 0 ) . h t t p s : / / d o i . o r g / 1 0 . 1 1 0 3 / scipostphyslectnotes.29
physrevphyseducres.16.020131 71. Dawid, A., et al.: Modern applications of machine learning in quantum
49. Aiello, C.D., et al.: Achieving a quantum smart workforce. Quantum Sci. sciences. arXiv preprint arXiv:2204.04198 (2022)
Technol. 6(3), 030501 (2021). [Link] 72. Pezzè, L., et al.: Machine learning for optical quantum metrology. Adv.
abfa64 Photon. 5(2), 020501 (2023). [Link]
50. Dzurak, A.S., et al.: Development of an undergraduate quantum engi- 73. Suter, D., Álvarez, G.A.: Colloquium: protecting quantum information
neering degree. IEEE Trans. Quantum Eng. 3, 1–10 (2022). https:// against environmental noise. Rev. Mod. Phys. 88(4), 041001 (2016).
[Link]/10.1109/tqe.2022.3157338 [Link]
51. Kaur, M., Venegas‐Gomez, A.: Defining the quantum workforce 74. Carrasquilla, J.: Machine learning for quantum matter. Adv. Phys. X 5(1),
landscape: a review of global quantum education initiatives. Opt. Eng. 1797528 (2020). [Link]
61(8), 081806 (2022). [Link] 75. Gebhart, V., et al.: Learning quantum systems. Nat. Rev. Phys. 5(3),
52. Ahmed, S.A., Mohsin, M., Ali, S.M.Z.: Survey and technological analysis 141–156 (2023). [Link]
of laser and its defense applications. Defence Technol. 17(2), 583–592 76. Zhu, Y., Yu, K.: Artificial intelligence (AI) for quantum and quantum
(2021). [Link] for AI. Opt. Quantum Electron. 55(8), 697 (2023). [Link]
53. Krelina, M.: Quantum technology for military applications. EPJ 1007/s11082‐023‐04914‐6
Quantum Technol. 8(1), 24 (2021). [Link] 77. Fontanesi, G., et al.: Artificial intelligence for satellite communication
s40507‐021‐00113‐y and non‐terrestrial networks: a survey. arXiv preprint arXiv:
54. Su, Y., Guo, Y., Huang, D.: Parameter optimization based BPNN of 2304.13008 (2023)
atmosphere continuous‐variable quantum key distribution. Entropy 78. Niemiec, M.: Error correction in quantum cryptography based on
21(9), 908 (2019). [Link] artificial neural networks. Quantum Inf. Process. 18(6), 174 (2019).
55. Wallnöfer, J., et al.: Machine learning for long‐distance quantum [Link]
communication. PRX Quantum 1(1), 010301 (2020). [Link] 79. Lohani, S., Glasser, R.T.: Turbulence correction with artificial neural
10.1103/prxquantum.1.010301 networks. Opt. Lett. 43(11), 2611–2614 (2018). [Link]
56. Li, Z., et al.: Ensemble learning for failure prediction of underwater 1364/ol.43.002611
continuous variable quantum key distribution with discrete modula- 80. You, C., et al.: Identification of light sources using machine learning.
tions. Phys. Lett. 419, 127694 (2021). [Link] Appl. Phys. Rev. 7(2) (2020). [Link]
physleta.2021.127694 81. Long, N.K., Malaney, R., Grant, K.J.: A survey of machine learning
57. Houssein, E.H., et al.: Machine learning in the quantum realm: the state‐ assisted continuous‐variable quantum key distribution. Information
of‐the‐art, challenges, and future vision. Expert Syst. Appl. 194, 116512 14(10), 553 (2023). [Link]
(2022). [Link] 82. Dunjko, V., Briegel, H.J.: Machine learning & artificial intelligence in the
58. Venketeswaran, A., et al.: Recent advances in machine learning for fiber quantum domain: a review of recent progress. Rep. Prog. Phys. 81(7),
optic sensor applications. Adv. Intell. Syst. 4(1), 2100067 (2022). 074001 (2018). [Link]
[Link] 83. Mabena, C.M., Roux, F.S.: Quantum channel correction with twisted
59. Krenn, M., et al.: Artificial intelligence and machine learning for light using compressive sensing. Phys. Rev. 101(1), 013807 (2020).
quantum technologies. Phys. Rev. 107(1), 010101 (2023). [Link] [Link]
org/10.1103/physreva.107.010101 84. Kundu, N.K., McKay, M.R., Mallik, R.K.: Machine‐learning‐based
60. Jin, D., et al.: Key‐sifting algorithms for continuous‐variable quantum parameter estimation of Gaussian quantum states. IEEE Trans. Quan-
key distribution. Phys. Rev. 104(1), 012616 (2021). [Link] tum Eng. 3, 1–13 (2021). [Link]
1103/physreva.104.012616 85. Cimini, V., et al.: Deep reinforcement learning for quantum multipa-
61. Rivest, R.L.: Cryptography and machine learning. In: International rameter estimation. Adv. Photon. 5(1), 016005 (2023). [Link]
Conference on the Theory and Application of Cryptology, pp. 427–439. 10.1117/[Link].5.1.016005
Springer (1991) 86. Turing, A.M.: Computing Machinery and Intelligence. Springer (2009)
 26328925, 2024, 3, Downloaded from [Link] by Pondicherry University, Wiley Online Library on [10/08/2025]. See the Terms and Conditions ([Link] on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
MAFU
- 225

87. Solomonoff, R.J.: The time scale of artificial intelligence: reflections on 113. Xu, D., Tian, Y.: A comprehensive survey of clustering algorithms. Ann.
social effects. Hum. Syst. Manag. 5(2), 149–153 (1985). [Link] Data Sci. 2, 165–193 (2015). [Link]
10.3233/hsm‐1985‐5207 0040‐1
88. Moor, J.: The Dartmouth college artificial intelligence conference: the 114. Hassoun, M.H.: Fundamentals of Artificial Neural Networks. MIT
next fifty years. AI Mag. 27(4), 87 (2006) Press (1995)
89. Rosch‐Grace, D., Straub, J.: Analysis of the likelihood of quantum 115. Jain, A.K., Mao, J., Mohiuddin, K.M.: Artificial neural networks: a tutorial.
computing proliferation. Technol. Soc. 68, 101880 (2022). [Link] Computer 29(3), 31–44 (1996). [Link]
org/10.1016/[Link].2022.101880 116. Gers, F.A., Schmidhuber, J., Cummins, F.: Learning to forget: continual
90. Weiss, G.: Multiagent Systems: A Modern Approach to Distributed prediction with LSTM. Neural Comput. 12(10), 2451–2471 (2000).
Artificial Intelligence. MIT Press (1999) [Link]
91. Janiesch, C., Zschech, P., Heinrich, K.: Machine learning and deep 117. Raymer, M.L., et al.: Dimensionality reduction using genetic algorithms.
learning. Electron. Mark. 31(3), 685–695 (2021). [Link] IEEE Trans. Evol. Comput. 4(2), 164–171 (2000). [Link]
1007/s12525‐021‐00475‐2 1109/4235.850656
92. Russell, S.J.: Artificial Intelligence a Modern Approach. Pearson Edu- 118. Dietterich, T.G.: Ensemble methods in machine learning. In: Interna-
cation, Inc. (2010) tional Workshop on Multiple Classifier Systems, pp. 1–15. Springer
93. Bishop, C.M., Nasrabadi, N.M.: Pattern Recognition and Machine (2000)
Learning, vol. 4. Springer (2006). [Link] 119. Zibar, D., et al.: Advancing classical and quantum communication sys-
45528‐0 tems with machine learning. In: Optical Fiber Communication Con-
94. Mehta, P., et al.: A high‐bias, low‐variance introduction to machine ference, pp. W1K–1. Optica Publishing Group (2020)
learning for physicists. Phys. Rep. 810, 1–124 (2019). [Link] 120. Sharma, P., et al.: Deep reinforcement learning‐based routing and
10.1016/[Link].2019.03.001 resource assignment in quantum key distribution‐secured optical net-
95. El Naqa, I., Murphy, M.J.: What Is Machine Learning? Springer (2015) works. IET Quantum Commun. 4(3), 136–145 (2023). [Link]
96. Muller, A.C., Guido, S.: Introduction to Machine Learning with Python. 10.1049/qtc2.12063
O’Reilly (2017) 121. Khan, F.N., et al.: An optical communication’s perspective on machine
97. Lake, B.M., et al.: Building machines that learn and think like people. learning and its applications. J. Lightwave Technol. 37(2), 493–516
Behav. Brain Sci. 40, e253 (2017). [Link] (2019). [Link]
s0140525x16001837 122. Saif, W.S., et al.: Machine learning techniques for optical performance
98. Goodfellow, I., Bengio, Y., Courville, A.: Deep Learning. MIT Press monitoring and modulation format identification: a survey. IEEE
(2016) Commun. Surv. Tutorials 22(4), 2839–2882 (2020). [Link]
99. Das Sarma, S., Deng, D.‐L., Duan, L.‐M.: Machine learning meets 1109/comst.2020.3018494
quantum physics. Phys. Today 72(3), 48–54 (2019). [Link] 123. Karanov, B., et al.: End‐to‐end deep learning of optical fiber commu-
1063/pt.3.4164 nications. J. Lightwave Technol. 36(20), 4843–4855 (2018). [Link]
100. Carleo, G., et al.: Machine learning and the physical sciences. Rev. Mod. org/10.1109/jlt.2018.2865109
Phys. 91(4), 045002 (2019). [Link] 124. Kleis, S., Schaeffer, C.G.: Improving the secret key rate of coherent
045002 quantum key distribution with Bayesian inference. J. Lightwave Technol.
101. Jiang, C., et al.: Machine learning paradigms for next‐generation wireless 37(3), 722–728 (2018). [Link]
networks. IEEE Wireless Commun. 24(2), 98–105 (2016). [Link] 125. Zibar, D., et al.: Machine learning techniques in optical communication.
org/10.1109/mwc.2016.1500356wc J. Lightwave Technol. 34(6), 1442–1452 (2015). [Link]
102. Merrick, K.E., Maher, M.L.: Motivated Reinforcement Learning: 1109/ecoc.2015.7341896
Curious Characters for Multiuser Games. Springer Science & Business 126. Ekert, A.K.: Quantum cryptography based on Bell’s theorem. Phys. Rev.
Media (2009) Lett. 67(6), 661–663 (1991). [Link]
103. Kaelbling, L.P., Littman, M.L., Moore, A.W.: Reinforcement learning: a 67.661
survey. J. Artif. Intell. Res. 4, 237–285 (1996). [Link] 127. Mafu, M., Senekane, M.: Security of quantum key distribution protocols.
jair.301 In: Advanced Technologies of Quantum Key Distribution. IntechOpen
104. LeCun, Y., Bengio, Y., Hinton, G.: Deep learning. Nature 521(7553), (2018)
436–444 (2015). [Link] 128. Karabo, K., et al.: A novel quantum key distribution resistant against
105. Ma, W., et al.: Deep learning for the design of photonic structures. Nat. large‐pulse attacks. IET Quantum Commun. (2024). [Link]
Photonics 15(2), 77–90 (2021). [Link] 10.1049/qtc2.12089
0685‐y 129. Liu, R., et al.: Towards the industrialisation of quantum key distribution
106. Kraus, M., Feuerriegel, S., Oztekin, A.: Deep learning in business ana- in communication networks: a short survey. IET Quantum Commun.
lytics and operations research: models, applications and managerial 3(3), 151–163 (2022). [Link]
implications. Eur. J. Oper. Res. 281(3), 628–641 (2020). [Link] 130. Busch, P., Lahti, P.J., Mittelstaedt, P.: The Quantum Theory of Mea-
10.1016/[Link].2019.09.018 surement. Springer (1996)
107. Cleveland, W.S., Devlin, S.J., Grosse, E.: Regression by local fitting: 131. Shannon, C.E.: Communication theory of secrecy systems. Bell Syst.
methods, properties, and computational algorithms. J. Econom. 37(1), Tech. J. 28(4), 656–715 (1949). [Link]
87–114 (1988). [Link] 1949.tb00928.x
108. Aha, D.W., Kibler, D., Albert, M.K.: Instance‐based learning algorithms. 132. Scarani, V., Kurtsiefer, C.: The black paper of quantum cryptography:
Mach. Learn. 6(1), 37–66 (1991). [Link] real implementation problems. Theor. Comput. Sci. 560, 27–32 (2014).
109. Bauer, F., Pereverzev, S., Rosasco, L.: On regularization algorithms in [Link]
learning theory. J. Complex 23(1), 52–72 (2007). [Link] 133. Diamanti, E., et al.: Practical challenges in quantum key distribution.
1016/[Link].2006.07.001 npj Quantum Inf. 2(1), 1–12 (2016). [Link]
110. Kotsiantis, S.B.: Decision trees: a recent overview. Artif. Intell. Rev. 2016.25
39(4), 261–283 (2013). [Link] 134. Brassard, G., et al.: Limitations on practical quantum cryptography.
111. Snoek, J., Larochelle, H., Adams, R.P.: Practical Bayesian optimization of Phys. Rev. Lett. 85(6), 1330–1333 (2000). [Link]
machine learning algorithms. Adv. Neural Inf. Process. Syst. 25 (2012) physrevlett.85.1330
112. Xu, R., Wunsch, D.: Survey of clustering algorithms. IEEE Trans. 135. Grosshans, F., Grangier, P.: Continuous variable quantum cryptography
Neural Netw. 16(3), 645–678 (2005). [Link] using coherent states. Phys. Rev. Lett. 88(5), 057902 (2002). [Link]
845141 org/10.1103/physrevlett.88.057902
 26328925, 2024, 3, Downloaded from [Link] by Pondicherry University, Wiley Online Library on [10/08/2025]. See the Terms and Conditions ([Link] on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
226
- MAFU

136. Lo, H.‐K., Ma, X., Chen, K.: Decoy state quantum key distribution. 158. Denys, A., Brown, P., Leverrier, A.: Explicit asymptotic secret key rate of
Phys. Rev. Lett. 94(23), 230504 (2005). [Link] continuous‐variable quantum key distribution with an arbitrary modulation.
9789812701633_0013 Quantum 5, 540 (2021). [Link]
137. Mayers, D., Yao, A.: Quantum cryptography with imperfect apparatus. 159. Lupo, C., Ouyang, Y.: Quantum key distribution with nonideal het-
In: Proceedings 39th Annual Symposium on Foundations of Computer erodyne detection: composable security of discrete‐modulation
Science (Cat. No. 98CB36280), pp. 503–509. IEEE (1998) continuous‐variable protocols. PRX Quantum 3(1), 010341 (2022).
138. Lucamarini, M., et al.: Overcoming the rate–distance limit of quantum [Link]
key distribution without quantum repeaters. Nature 557(7705), 400–403 160. Ding, C., et al.: Machine‐learning‐based detection for quantum hacking
(2018). [Link] attacks on continuous‐variable quantum‐key‐distribution systems. Phys.
139. Portmann, C., Renner, R.: Security in quantum cryptography. Rev. Mod. Rev. 107(6), 062422 (2023). [Link]
Phys. 94(2), 025008 (2022). [Link] 062422
025008 161. Jain, N., et al.: Device calibration impacts security of quantum key
140. Ekert, A., Renner, R.: The ultimate physical limits of privacy. Nature distribution. Phys. Rev. Lett. 107(11), 110501 (2011). [Link]
507(7493), 443–447 (2014). [Link] 10.1103/physrevlett.107.110501
141. Sekga, C., Mafu, M., Senekane, M.: High‐dimensional quantum key 162. Makarov, V., Anisimov, A., Skaar, J.: Effects of detector efficiency
distribution implemented with biphotons. Sci. Rep. 13(1), 1229 (2023). mismatch on security of quantum cryptosystems. Phys. Rev. 74(2),
[Link] 022313 (2006). [Link]
142. Sekga, C., Mafu, M.: Tripartite quantum key distribution implemented 163. Fung, C.‐H.F., et al.: Security proof of quantum key distribution with
with imperfect sources. Optics 3(3), 191–208 (2022). [Link] detection efficiency mismatch. arXiv preprint arXiv:0802.3788 (2008)
10.3390/opt3030019 164. Zhang, Y., et al.: Security proof of practical quantum key distribution
143. Garapo, K., Mafu, M., Petruccione, F.: Intercept‐resend attack on six‐ with detection‐efficiency mismatch. Phys. Rev. Res. 3(1), 013076 (2021).
state quantum key distribution over collective‐rotation noise channels. [Link]
Chin. Phys. B 25(7), 070303 (2016). [Link] 165. Lo, H.‐K., Curty, M., Qi, B.: Measurement‐device‐independent quan-
1056/25/7/070303 tum key distribution. Phys. Rev. Lett. 108(13), 130503 (2012). https://
144. Mafu, M., Garapo, K., Petruccione, F.: Finite‐key‐size security of the [Link]/10.1103/physrevlett.108.130503
Phoenix‐Barnett‐Chefles 2000 quantum‐key‐distribution protocol. 166. Langenfeld, S., et al.: Quantum repeater node demonstrating uncondi-
Phys. Rev. 90(3), 032308 (2014). [Link] tionally secure key distribution. Phys. Rev. Lett. 126(23), 230506 (2021).
032308 [Link]
145. Lütkenhaus, N.: Security against individual attacks for realistic quantum 167. Ren, Z.‐A., et al.: Implementation of machine learning in quantum key
key distribution. Phys. Rev. 61(5), 052304 (2000). [Link] distributions. IEEE Commun. Lett. 25(3), 940–944 (2020). [Link]
1103/physreva.61.052304 org/10.1109/lcomm.2020.3040212
146. Matsuura, T., et al.: Finite‐size security of continuous‐variable quantum 168. Tang, G.‐Z., Li, C.‐Y., Wang, M.: Polarization discriminated time‐bin
key distribution with digital signal processing. Nat. Commun. 12(1), 252 phase‐encoding measurement‐device‐independent quantum key distri-
(2021). [Link] bution. Quantum Eng. 3(4), e79 (2021). [Link]
147. Sekga, C., Mafu, M.: Security of quantum‐key‐distribution protocol by que2.79
using the post‐selection technique. Phys. Open 7, 100075 (2021). 169. Okey, O.D., et al.: Quantum key distribution protocol selector based on
[Link] machine learning for next‐generation networks. Sustainability 14(23),
148. Coles, P.J., Metodiev, E.M., Lütkenhaus, N.: Numerical approach for 15901 (2022). [Link]
unstructured quantum key distribution. Nat. Commun. 7(1), 11712 170. Hu, H., et al.: Robust interior point method for quantum key distri-
(2016). [Link] bution rate computation. Quantum 6, 792 (2022). [Link]
149. Winick, A., Lütkenhaus, N., Coles, P.J.: Reliable numerical key rates for 22331/q‐2022‐09‐08‐792
quantum key distribution. Quantum 2, 77 (2018). [Link] 171. Liu, Z.‐P., et al.: Automated machine learning for secure key rate in
22331/q‐2018‐07‐26‐77 discrete‐modulated continuous‐variable quantum key distribution. Opt.
150. Lin, J., Lütkenhaus, N.: Trusted detector noise analysis for discrete Express 30(9), 15024–15036 (2022). [Link]
modulation schemes of continuous‐variable quantum key distribution. 172. Zhou, M.‐G., et al.: Neural network‐based prediction of the secret‐key
Phys. Rev. Appl. 14(6), 064030 (2020). [Link] rate of quantum key distribution. Sci. Rep. 12(1), 8879 (2022). https://
physrevapplied.14.064030 [Link]/10.1038/s41598‐022‐12647‐x
151. Tan, E.Y.‐Z., et al.: Computing secure key rates for quantum cryptog- 173. George, I., Lin, J., Lütkenhaus, N.: Numerical calculations of the finite
raphy with untrusted devices. npj Quantum Inf. 7(1), 158 (2021). key rate for general quantum key distribution protocols. Phys. Rev. Res.
[Link] 3(1), 013274 (2021). [Link]
152. Gottesman, D., et al.: Security of quantum key distribution with 174. Dong, Q., et al.: Optimization parameter prediction‐based XGBoost of
imperfect devices. In: International Symposium onInformation Theory, TF‐QKD. Quantum Inf. Process. 21(7), 233 (2022). [Link]
2004. ISIT 2004. Proceedings, pp. 136. IEEE (2004) 1007/s11128‐022‐03579‐6
153. Sekga, C., Mafu, M.: Reference frame independent twin field quantum 175. Liu, W., et al.: Monitoring of continuous‐variable quantum key distri-
key distribution with source flaws. J. Phys. Commun. 5(4), 045008 bution system in real environment. Opt Express 25(16), 19429–19443
(2021). [Link] (2017). [Link]
154. Shor, P.W., Preskill, J.: Simple proof of security of the BB84 quantum 176. Liu, W., et al.: Integrating machine learning to achieve an automatic
key distribution protocol. Phys. Rev. Lett. 85(2), 441–444 (2000). parameter prediction for practical continuous‐variable quantum key
[Link] distribution. Phys. Rev. 97(2), 022316 (2018). [Link]
155. Briegel, H.‐J., et al.: Quantum repeaters: the role of imperfect local physreva.97.022316
operations in quantum communication. Phys. Rev. Lett. 81(26), 177. Li, J., et al.: Discrete‐modulated continuous‐variable quantum key dis-
5932–5935 (1998). [Link] tribution with a machine‐learning‐based detector. Opt. Eng. 57(6),
156. Sangouard, N., et al.: Quantum repeaters based on atomic ensembles 066109 (2018). [Link]
and linear optics. Rev. Mod. Phys. 83(1), 33–80 (2011). [Link] 178. Wang, W., Lo, H.‐K.: Machine learning for optimal parameter prediction
10.1103/revmodphys.83.33 in quantum key distribution. Phys. Rev. 100(6), 062334 (2019). https://
157. Pirandola, S.: End‐to‐end capacities of a quantum communication [Link]/10.1103/physreva.100.062334
network. Commun. Phys. 2(1), 51 (2019). [Link] 179. Scarani, V., Renner, R.: Quantum cryptography with finite resources:
s42005‐019‐0147‐3 unconditional security bound for discrete‐variable protocols with one‐
 26328925, 2024, 3, Downloaded from [Link] by Pondicherry University, Wiley Online Library on [10/08/2025]. See the Terms and Conditions ([Link] on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
MAFU
- 227

way postprocessing. Phys. Rev. Lett. 100(20), 200501 (2008). https:// 201. Nölleke, C., et al.: Efficient teleportation between remote single‐atom
[Link]/10.1103/physrevlett.100.200501 quantum memories. Phys. Rev. Lett. 110(14), 140403 (2013). https://
180. Mafu, M., Garapo, K., Petruccione, F.: Finite‐size key in the Bennett 1992 [Link]/10.1103/physrevlett.110.140403
quantum‐key‐distribution protocol for Rényi entropies. Phys. Rev. A 202. Flamini, F., Spagnolo, N., Sciarrino, F.: Photonic quantum information
88(6), 062306 (2013). [Link] processing: a review. Rep. Prog. Phys. 82(1), 016001 (2018). [Link]
181. Ding, H.‐J., et al.: Predicting optimal parameters with random forest for org/10.1088/1361‐6633/aad5b2
quantum key distribution. Quantum Inf. Process. 19(2), 1–8 (2020). 203. Kim, Y.‐H., Kulik, S.P., Shih, Y.: Quantum teleportation of a polariza-
[Link] tion state with a complete Bell state measurement. Phys. Rev. Lett. 86(7),
182. Lu, F.‐Y., et al.: Parameter optimization and real‐time calibration of a 1370–1373 (2001). [Link]
measurement‐device‐independent quantum key distribution network 204. Marcikic, I., et al.: Long‐distance teleportation of qubits at telecom-
based on a back propagation artificial neural network. JOSA B 36(3), munication wavelengths. Nature 421(6922), 509–513 (2003). https://
B92–B98 (2019). [Link] [Link]/10.1038/nature01376
183. Liu, J.‐Y., et al.: Practical phase‐modulation stabilization in quantum key 205. Landry, O., et al.: Quantum teleportation over the Swisscom telecom-
distribution via machine learning. Phys. Rev. Appl. 12(1), 014059 (2019). munication network. JOSA B 24(2), 398–403 (2007). [Link]
[Link] 10.1364/josab.24.000398
184. Mao, Y., et al.: Detecting quantum attacks: a machine learning based 206. Bao, X.‐H., et al.: Quantum teleportation between remote atomic‐
defense strategy for practical continuous‐variable quantum key distri- ensemble quantum memories. Proc. Natl. Acad. Sci. USA 109(50),
bution. New J. Phys. 22(8), 083073 (2020). [Link] 20347–20351 (2012). [Link]
1367‐2630/aba8d4 207. Reindl, M., et al.: All‐photonic quantum teleportation using on‐demand
185. Zhang, S., et al.: Machine learning‐assisted measurement device‐ solid‐state quantum emitters. Sci. Adv. 4(12), eaau1255 (2018). https://
independent quantum key distribution on reference frame calibration. [Link]/10.1126/sciadv.aau1255
Entropy 23(10), 1242 (2021). [Link] 208. Gao, W., et al.: Quantum teleportation from a propagating photon to a
186. Xu, F., Xu, H., Lo, H.‐K.: Protocol choice and parameter optimization solid‐state spin qubit. Nat. Commun. 4(1), 2744 (2013). [Link]
in decoy‐state measurement‐device‐independent quantum key distri- org/10.1038/ncomms3744
bution. Phys. Rev. 89(5), 052333 (2014). [Link] 209. Nielsen, M.A., Knill, E., Laflamme, R.: Complete quantum teleportation
physreva.89.052333 using nuclear magnetic resonance. Nature 396(6706), 52–55 (1998).
187. Ismail, Y., Sinayskiy, I., Petruccione, F.: Integrating machine learning [Link]
techniques in quantum communication to characterize the quantum 210. Huo, M., et al.: Deterministic quantum teleportation through fiber
channel. JOSA B 36(3), B116–B121 (2019). [Link] channels. Sci. Adv. 4(10), eaas9401 (2018). [Link]
josab.36.00b116 sciadv.aas9401
188. Lollie, M.L., et al.: High‐dimensional encryption in optical fibers using 211. Furusawa, A., et al.: Unconditional quantum teleportation. Science
spatial modes of light and machine learning. Mach. Learn. Sci. Technol. 282(5389), 706–709 (1998). [Link]
3(3), 035006 (2022). [Link] 5389.706
189. Ou, Y., et al.: Field‐trial of machine learning‐assisted quantum key 212. Ursin, R., et al.: Quantum teleportation across the Danube. Nature
distribution (QKD) networking with SDN. In: 2018 European Con- 430(7002), 849 (2004). [Link]
ference on Optical Communication (ECOC), pp. 1–3. IEEE (2018) 213. Valivarthi, R., et al.: Teleportation systems toward a quantum internet.
190. Khan, F.N., Lu, C., Lau, A.P.T.: Machine learning methods for optical PRX Quantum 1(2), 020317 (2020). [Link]
communication systems. In: Signal Processing in Photonic Communi- prxquantum.1.020317
cations, pp. SpW2F–3. Optica Publishing Group (2017) 214. Takesue, H., et al.: Quantum teleportation over 100 km of fiber using
191. Sekga, C., Mafu, M.: Three‐party reference frame independent quantum highly efficient superconducting nanowire single‐photon detectors.
key distribution protocol. Chin. Phys. B 30(12), 120301 (2021). https:// Optica 2(10), 832–835 (2015). [Link]
[Link]/10.1088/1674‐1056/abff1f 215. Zhao, H., et al.: Real time deterministic quantum teleportation over 10
192. Liao, Q., et al.: Multi‐label learning for improving discretely‐modulated km of single optical fiber channel. Opt. Express 30(3), 3770–3782
continuous‐variable quantum key distribution. New J. Phys. 22(8), (2022). [Link]
083086 (2020). [Link] 216. Wang, X.‐L., et al.: Quantum teleportation of multiple degrees of
193. Zhu, Y., et al.: Passive continuous variable quantum key distribution freedom of a single photon. Nature 518(7540), 516–519 (2015). https://
through the oceanic turbulence. Entropy 25(2), 307 (2023). [Link] [Link]/10.1038/nature14246
org/10.3390/e25020307 217. Riebe, M., et al.: Deterministic quantum teleportation with atoms.
194. Huang, D., Liu, S., Zhang, L.: Secure continuous‐variable quantum key Nature 429(6993), 734–737 (2004). [Link]
distribution with machine learning. In: Photonics, vol. 8(11), p. 511. nature02570
MDPI (2021). [Link] 218. Barrett, M., et al.: Deterministic quantum teleportation of atomic qubits.
195. Bouwmeester, D., et al.: Experimental quantum teleportation. Nature Nature 429(6993), 737–739 (2004). [Link]
390(6660), 575–579 (1997). [Link] nature02608
196. Braunstein, S.L., Kimble, H.J.: Teleportation of continuous quantum 219. Häffner, H., Roos, C.F., Blatt, R.: Quantum computing with trapped
variables. Phys. Rev. Lett. 80(4), 869–872 (1998). [Link] ions. Phys. Rep. 469(4), 155–203 (2008). [Link]
1103/physrevlett.80.869 physrep.2008.09.003
197. Bell, J.S.: On the problem of hidden variables in quantum mechanics. 220. Yin, J., et al.: Quantum teleportation and entanglement distribution over
Rev. Mod. Phys. 38(3), 447–452 (1966). [Link] 100‐kilometre free‐space channels. Nature 488(7410), 185–188 (2012).
revmodphys.38.447 [Link]
198. Reid, M., et al.: Colloquium: the Einstein‐Podolsky‐Rosen paradox: 221. Li, B., et al.: Quantum state transfer over 1200 km assisted by prior
from concepts to applications. Rev. Mod. Phys. 81(4), 1727–1751 distributed entanglement. Phys. Rev. Lett. 128(17), 170501 (2022).
(2009). [Link] [Link]
199. Pirandola, S., et al.: Advances in quantum teleportation. Nat. Photonics 222. Ren, J.‐G., et al.: Ground‐to‐satellite quantum teleportation. Nature
9(10), 641–652 (2015). [Link] 549(7670), 70–73 (2017). [Link]
200. Bruzewicz, C.D., et al.: Trapped‐ion quantum computing: progress and 223. Ma, X.‐S., et al.: Quantum teleportation over 143 kilometres using active
challenges. Appl. Phys. Rev. 6(2) (2019). [Link] feed‐forward. Nature 489(7415), 269–273 (2012). [Link]
5088164 1038/nature11472
 26328925, 2024, 3, Downloaded from [Link] by Pondicherry University, Wiley Online Library on [10/08/2025]. See the Terms and Conditions ([Link] on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
228
- MAFU

224. Bennett, C.H., et al.: Purification of noisy entanglement and faithful 246. Erhard, M., Krenn, M., Zeilinger, A.: Advances in high‐dimensional
teleportation via noisy channels. Phys. Rev. Lett. 76(5), 722–725 (1996). quantum entanglement. Nat. Rev. Phys. 2(7), 365–381 (2020). https://
[Link] [Link]/10.1038/s42254‐020‐0193‐5
225. Xia, X.‐X., et al.: Long distance quantum teleportation. Quantum Sci. 247. McCutcheon, W., et al.: Experimental verification of multipartite
Technol. 3(1), 014012 (2017). [Link] entanglement in quantum networks. Nat. Commun. 7(1), 13251 (2016).
aa9baf [Link]
226. Simon, C.: Towards a global quantum network. Nat. Photonics 11(11), 248. Calsamiglia, J.: Generalized measurements by linear elements. Phys. Rev.
678–680 (2017). [Link] 65(3), 030301 (2002). [Link]
227. Cacciapuoti, A.S., et al.: When entanglement meets classical communi- 249. Zhong, H.‐S., et al.: 12‐photon entanglement and scalable scattershot
cations: quantum teleportation for the quantum internet. IEEE Trans. boson sampling with optimal entangled‐photon pairs from parametric
Commun. 68(6), 3808–3833 (2020). [Link] down‐conversion. Phys. Rev. Lett. 121(25), 250505 (2018). [Link]
2020.2978071 org/10.1103/physrevlett.121.250505
228. Cao, Y., et al.: The evolution of quantum key distribution networks: on 250. Qiu, X., et al.: Optical vortex copier and regenerator in the Fourier
the road to the Qinternet. IEEE Commun. Surv. Tutorials 24(2), domain. Photon. Res. 6(6), 641–646 (2018). [Link]
839–894 (2022). [Link] prj.6.000641
229. Zukowski, M., et al.: “Event‐ready‐detectors” Bell experiment via 251. Kaneda, F., Kwiat, P.G.: High‐efficiency single‐photon generation via
entanglement swapping. Phys. Rev. Lett. 71(26), 4287–4290 (1993). large‐scale active time multiplexing. Sci. Adv. 5(10), eaaw8586 (2019).
[Link] [Link]
230. Pirandola, S., et al.: Macroscopic entanglement by entanglement swap- 252. Schimpf, C., et al.: Quantum dots as potential sources of strongly
ping. Phys. Rev. Lett. 97(15), 150403 (2006). [Link] entangled photons: perspectives and challenges for applications in
physrevlett.97.150403 quantum networks. Appl. Phys. Lett. 118(10) (2021). [Link]
231. Takeda, S., et al.: Entanglement swapping between discrete and 10.1063/5.0038729
continuous variables. Phys. Rev. Lett. 114(10), 100501 (2015). https:// 253. Lu, J., et al.: Ultralow‐threshold thin‐film lithium niobate optical para-
[Link]/10.1103/physrevlett.114.100501 metric oscillator. Optica 8(4), 539–544 (2021). [Link]
232. Wang, T.‐J., Yang, G.‐Q., Wang, C.: Control power of high‐dimensional optica.418984
controlled teleportation. Phys. Rev. 101(1), 012323 (2020). [Link] 254. Ringbauer, M., et al.: A universal qudit quantum processor with trapped
org/10.1103/physreva.101.012323 ions. Nat. Phys. 18(9), 1053–1057 (2022). [Link]
233. Barasiński, A., Černoch, A., Lemr, K.: Demonstration of controlled s41567‐022‐01658‐0
quantum teleportation for discrete variables on linear optical devices. 255. Fu, Y., et al.: Experimental investigation of quantum correlations in a
Phys. Rev. Lett. 122(17), 170501 (2019). [Link] two‐qutrit spin system. Phys. Rev. Lett. 129(10), 100501 (2022). https://
physrevlett.122.170501 [Link]/10.1103/physrevlett.129.100501
234. Lance, A.M., et al.: Tripartite quantum state sharing. Phys. Rev. 256. Goss, N., et al.: High‐fidelity qutrit entangling gates for superconducting
Lett. 92(17), 177903 (2004). [Link] circuits. Nat. Commun. 13(1), 7481 (2022). [Link]
177903 s41467‐022‐34851‐z
235. Bužek, V., Hillery, M.: Quantum copying: beyond the no‐cloning the- 257. Xu, J., et al.: Machine learning assisted prediction for free‐space
orem. Phys. Rev. 54(3), 1844–1852 (1996). [Link] continuous variable quantum teleportation. IEEE Photon. J. 14(4),
physreva.54.1844 1–7 (2022). [Link]
236. Yonezawa, H., Aoki, T., Furusawa, A.: Demonstration of a quantum 258. Shamir, A.: How to share a secret. Commun. ACM 22(11), 612–613
teleportation network for continuous variables. Nature 431(7007), (1979). [Link]
430–433 (2004). [Link] 259. Blakley, G.R.: Safeguarding cryptographic keys. In: Managing Re-
237. Ishizaka, S., Hiroshima, T.: Asymptotic teleportation scheme as a uni- quirements Knowledge, International Workshop on, pp. 313. IEEE
versal programmable quantum processor. Phys. Rev. Lett. 101(24), 240501 Computer Society (1979)
(2008). [Link] 260. Cleve, R., Gottesman, D., Lo, H.‐K.: How to share a quantum secret.
238. Ishizaka, S., Hiroshima, T.: Quantum teleportation scheme by selecting Phys. Rev. Lett. 83(3), 648–651 (1999). [Link]
one of multiple output ports. Phys. Rev. 79(4), 042306 (2009). https:// physrevlett.83.648
[Link]/10.1103/physreva.79.042306 261. Sekga, C., Mafu, M.: Quantum state sharing of an arbitrary m‐particle
239. Gottesman, D., Chuang, I.L.: Demonstrating the viability of uni- state using Einstein–Podolsky–Rosen pairs and application in quan-
versal quantum computation using teleportation and single‐qubit tum voting. Mod. Phys. Lett. 36(21), 2150151 (2021). [Link]
operations. Nature 402(6760), 390–393 (1999). [Link] 10.1142/s0217732321501510
1038/46503 262. Al Ebri, N., Baek, J., Yeun, C.Y.: Study on secret sharing schemes
240. Aliferis, P., Leung, D.W.: Computation by measurements: a unifying (SSS) and their applications. In: 2011 International Conference for
picture. Phys. Rev. 70(6), 062314 (2004). [Link] Internet Technology and Secured Transactions, pp. 40–45. IEEE
physreva.70.062314 (2011)
241. Lu, S., et al.: Separability‐entanglement classifier via machine learning. 263. Chor, B., et al.: Verifiable secret sharing and achieving simultaneity in
Phys. Rev. 98(1), 012315 (2018). [Link] the presence of faults. In: 26th Annual Symposium on Foundations of
012315 Computer Science (SFCS 1985), pp. 383–395. IEEE (1985)
242. Asif, N., et al.: Entanglement detection with artificial neural networks. 264. Beimel, A.: Secret‐sharing schemes: a survey. In: International Confer-
Sci. Rep. 13(1), 1562 (2023). [Link] ence on Coding and Cryptology, pp. 11–46. Springer (2011)
28745‐3 265. Cramer, R., Damgård, I., Maurer, U.: General secure multi‐party
243. Boerkamp, M.: Quantum computer creates wormhole analogue. Phys. computation from any linear secret‐sharing scheme. In: International
World 36(1), 6 (2023). [Link] Conference on the Theory and Applications of Cryptographic Tech-
244. Luo, Y.‐H., et al.: Quantum teleportation in high dimensions. Phys. Rev. niques, pp. 316–334. Springer (2000)
Lett. 123(7), 070505 (2019). [Link] 266. Cramer, R., Damgård, I.B.: Secure Multiparty Computation. Cambridge
070505 University Press (2015)
245. Hu, X.‐M., et al.: Experimental high‐dimensional quantum teleporta- 267. Goyal, R., Vaikuntanathan, V.: Locally verifiable signature and key ag-
tion. Phys. Rev. Lett. 125(23), 230501 (2020). [Link] gregation. In: Annual International Cryptology Conference, pp. 761–791.
physrevlett.125.230501 Springer (2022)
 26328925, 2024, 3, Downloaded from [Link] by Pondicherry University, Wiley Online Library on [10/08/2025]. See the Terms and Conditions ([Link] on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
MAFU
- 229

268. Bellare, M., Neven, G.: Identity‐based multi‐signatures from RSA. In: 290. Guo, G.‐P., Guo, G.‐C.: Quantum secret sharing without entanglement.
Cryptographers’ Track at the RSA Conference, pp. 145–162. Springer Phys. Lett. 310(4), 247–251 (2003). [Link]
(2007) 9601(03)00074‐4
269. Boldyreva, A., et al.: Ordered multisignatures and identity‐based 291. Wang, M.‐M., Tian, L.‐T., Qu, Z.‐G.: Efficient multiparty quantum
sequential aggregate signatures, with applications to secure routing. In: secret sharing scheme in high‐dimensional system. In: International
Proceedings of the 14th ACM Conference on Computer and Com- Conference on Cloud Computing and Security, pp. 23–31. Springer
munications Security, pp. 276–285 (2007) (2018)
270. Goyal, V., et al.: Attribute‐based encryption for fine‐grained access 292. Gao, M., Yang, W., Liu, Y.: A novel quantum (t, n) threshold group
control of encrypted data. In: Proceedings of the 13th ACM Conference signature based on d‐dimensional quantum system. Quantum Inf.
on Computer and Communications Security, pp. 89–98 (2006) Process. 20, 1–13 (2021)
271. Bethencourt, J., Sahai, A., Waters, B.: Ciphertext‐policy attribute‐based 293. Hu, W., et al.: A novel dynamic quantum secret sharing in high‐
encryption. In: 2007 IEEE Symposium on Security and Privacy dimensional quantum system. Quantum Inf. Process. 20(5), 1–28
(SP’07), pp. 321–334. IEEE (2007) (2021). [Link]
272. Gottesman, D.: Theory of quantum secret sharing. Phys. Rev. 61(4), 294. Pinnell, J., et al.: Experimental demonstration of 11‐dimensional 10‐
042311 (2000). [Link] party quantum secret sharing. Laser Photon. Rev. 14(9), 2000012
273. Zhang, Z.‐J., Li, Y., Man, Z.‐X.: Multiparty quantum secret sharing. (2020). [Link]
Phys. Rev. 71(4), 044301 (2005). [Link] 295. Hai‐Qiang, M., Ke‐Jin, W., Jian‐Hui, Y.: Experimental single qubit
044301 quantum secret sharing in a fiber network configuration. Opt. Lett.
274. Tavakoli, A., et al.: Secret sharing with a single d‐level quantum system. 38(21), 4494–4497 (2013). [Link]
Phys. Rev. 92(3), 030302 (2015). [Link] 296. Choi, I., et al.: Field trial of a quantum secured 10 Gb/s DWDM
030302 transmission system over a single installed fiber. Opt. Express 22(19),
275. Liu, L.‐J., et al.: A quantum secret sharing scheme with verifiable 23121–23128 (2014). [Link]
function. Eur. Phys. J. D 74(7), 1–8 (2020). [Link] 297. Wiesner, S.: Conjugate coding. ACM Sigact News 15(1), 78–88 (1983).
epjd/e2020‐10010‐3 [Link]
276. Li, F., Yan, J., Zhu, S.: General quantum secret sharing scheme based on 298. Yin, H.‐L., et al.: Experimental quantum secure network with digital
two qudit. Quantum Inf. Process. 20(10), 328 (2021). [Link] signatures and encryption. Natl. Sci. Rev. 10(4), nwac228 (2023).
10.1007/s11128‐021‐03270‐2 [Link]
277. Hsu, L.‐Y., Li, C.‐M.: Quantum secret sharing using product states. Phys. 299. Chen, C.‐Y., et al.: Quantum cryptography and its applications over the
Rev. 71(2), 022321 (2005). [Link] internet. IEEE Netw. 29(5), 64–69 (2015). [Link]
278. Mashhadi, S.: General secret sharing based on quantum Fourier trans- mnet.2015.7293307
form. Quantum Inf. Process. 18(4), 114 (2019). [Link] 300. Tian, J.‐H., Zhang, J.‐Z., Li, Y.‐P.: A voting protocol based on the
1007/s11128‐019‐2233‐6 controlled quantum operation teleportation. Int. J. Theor. Phys. 55(5),
279. Tsai, C., Hwang, T.: Multi‐party quantum secret sharing based on two 2303–2310 (2016). [Link]
special entangled states. Sci. China Phys. Mech. Astron. 55(3), 460–464 301. Tanuwidjaja, H.C., et al.: Privacy‐preserving deep learning on machine
(2012). [Link] learning as a service—a comprehensive survey. IEEE Access 8,
280. Chong‐Qiang, Y., et al.: Multiparty semi‐quantum secret sharing with d‐ 167425–167447 (2020). [Link]
level single‐particle states. Int. J. Theor. Phys. 58(11), 3797–3814 (2019). 302. Lindell, Y., Pinkas, B.: Privacy preserving data mining. J. Cryptol. 15(3),
[Link] 177–206 (2002). [Link]
281. Shi, R., et al.: Quantum secret sharing between multiparty and multi- 303. Senekane, M., Mafu, M., Taele, B.M.: Privacy‐preserving quantum ma-
party with Bell states and Bell measurements. Sci. China Phys. Mech. chine learning using differential privacy. In: 2017 IEEE AFRICON, pp.
Astron. 53(12), 2238–2244 (2010). [Link] 1432–1435. IEEE (2017)
010‐4181‐0 304. Yao, A.C.: Protocols for secure computations. In: 23rd Annual Sympo-
282. Zhou, R.‐G., et al.: Dynamic multiparty quantum secret sharing with a sium on Foundations of Computer Science (SFCS 1982), pp. 160–164.
trusted party based on generalized GHZ state. IEEE Access 9, IEEE (1982)
22986–22995 (2021). [Link] 305. Rivest, R.L., et al.: On data banks and privacy homomorphisms. Found.
283. Yang, Y., Wen, Q.: Threshold quantum secret sharing between multi‐ Secure Comput. 4(11), 169–180 (1978)
party and multi‐party. Sci China G Phys. Mech. Astron. 51(9), 306. Dwork, C.: Differential privacy. In: International Colloquium on
1308–1315 (2008). [Link] Automata, Languages, and Programming, pp. 1–12. Springer (2006)
284. Senthoor, K., Sarvepalli, P.K.: Communication efficient quantum secret 307. Dong, Y., et al.: Privacy‐preserving distributed machine learning based
sharing. Phys. Rev. 100(5), 052313 (2019). [Link] on secret sharing. In: Information and Communications Security: 21st
physreva.100.052313 International Conference, ICICS 2019, Beijing, China, December 15–
285. Wang, M.‐M., Chen, X.‐B., Yang, Y.‐X.: Quantum secret sharing for 17, 2019, Revised Selected Papers 21, pp. 684–702. Springer (2020)
general access structures based on multiparticle entanglements. Quan- 308. Wei, W., Tang, C., Chen, Y.: Efficient privacy‐preserving K‐means
tum Inf. Process. 13(2), 429–443 (2014). [Link] clustering from secret‐sharing‐based secure three‐party computation.
s11128‐013‐0660‐3 Entropy 24(8), 1145 (2022). [Link]
286. Deng, F.‐G., Zhou, H.‐Y., Long, G.L.: Circular quantum secret sharing. 309. Araki, T., et al.: High‐throughput semi‐honest secure three‐party
J. Phys. Math. Gen. 39(45), 14089–14099 (2006). [Link] computation with an honest majority. In: Proceedings of the 2016
1088/0305‐4470/39/45/018 ACM SIGSAC Conference on Computer and Communications Secu-
287. Yang, C.‐W., Tsai, C.‐W.: Efficient and secure dynamic quantum secret rity, pp. 805–817 (2016)
sharing protocol based on Bell states. Quantum Inf. Process. 19(5), 1–14 310. Hegde, A., et al.: SoK: efficient privacy‐preserving clustering. Cryp-
(2020). [Link] tology ePrint Archive (2021)
288. Wang, T.‐Y., et al.: Analysis of efficient and secure dynamic quantum 311. Kozlowski, W., Dahlberg, A., Wehner, S.: Designing a quantum network
secret sharing protocol based on Bell states. Quantum Inf. Process. 20, protocol. In: Proceedings of the 16th International Conference on
1–10 (2021). [Link] Emerging Networking Experiments and Technologies, pp. 1–16 (2020)
289. Tittel, W., Zbinden, H., Gisin, N.: Experimental demonstration of 312. Van Meter, R., Touch, J.: Designing quantum repeater networks. IEEE
quantum secret sharing. Phys. Rev. 63(4), 042301 (2001). [Link] Commun. Mag. 51(8), 64–71 (2013). [Link]
org/10.1103/physreva.63.042301 2013.6576340
 26328925, 2024, 3, Downloaded from [Link] by Pondicherry University, Wiley Online Library on [10/08/2025]. See the Terms and Conditions ([Link] on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
230
- MAFU

313. Van Meter, R., et al.: A quantum internet architecture. In: 2022 IEEE 335. Xia, Y., et al.: Quantum‐enhanced data classification with a variational
International Conference on Quantum Computing and Engineering entangled sensor network. Phys. Rev. X 11(2), 021047 (2021). https://
(QCE), pp. 341–352. IEEE (2022) [Link]/10.1103/physrevx.11.021047
314. Wei, S.‐H., et al.: Towards real‐world quantum networks: a review. Laser 336. Pant, M., et al.: Routing entanglement in the quantum internet. npj Quan-
Photon. Rev. 16(3), 2100219 (2022). [Link] tum Inf. 5(1), 25 (2019). [Link]
202100219 337. Coutinho, B.C., et al.: Robustness of noisy quantum networks. Com-
315. Kozlowski, W., Wehner, S.: Towards large‐scale quantum networks. In: mun. Phys. 5(1), 105 (2022). [Link]
Proceedings of the Sixth Annual ACM International Conference on 00866‐7
Nanoscale Computing and Communication, pp. 1–7 (2019) 338. Helsen, J., Wehner, S.: A benchmarking procedure for quantum net-
316. Cao, Y., et al.: Multi‐tenant secret‐key assignment over quantum key works. npj Quantum Inf. 9(1), 17 (2023). [Link]
distribution networks. Opt. Express 27(3), 2544–2561 (2019). https:// s41534‐022‐00628‐x
[Link]/10.1364/oe.27.002544 339. Amirloo, J., Razavi, M., Majedi, A.H.: DLCZ quantum repeaters: rate
317. Wang, J., et al.: Integrated photonic quantum technologies. Nat. Pho- and fidelity analysis. In: International Conference on Quantum Infor-
tonics 14(5), 273–284 (2020). [Link] mation, p. JMB54. Optica Publishing Group (2008)
0532‐1 340. Razavi, M., Farmanbar, H., Lütkenhaus, N.: Long‐distance quantum
318. O’brien, J.L., Furusawa, A., Vučković, J.: Photonic quantum technolo- communication with multiple quantum memories. In: Optical Fiber
gies. Nat. Photonics 3(12), 687–695 (2009). [Link] Communication Conference, p. JWA48. Optica Publishing Group
nphoton.2009.229 (2008)
319. Kwek, L.‐C., et al.: Chip‐based quantum key distribution. AAPPS Bull. 341. Yin, J., et al.: Satellite‐based entanglement distribution over 1200 kilo-
31, 1–8 (2021). [Link] meters. Science 356(6343), 1140–1144 (2017). [Link]
320. Sibson, P., et al.: Integrated silicon photonics for high‐speed quantum science.aan3211
key distribution. Optica 4(2), 172–177 (2017). [Link] 342. Yin, J., et al.: Satellite‐to‐ground entanglement‐based quantum key
optica.4.000172 distribution. Phys. Rev. Lett. 119(20), 200501 (2017). [Link]
321. Ding, Y., et al.: High‐dimensional quantum key distribution based on 10.1103/physrevlett.119.200501
multicore fiber using silicon photonic integrated circuits. npj Quantum 343. Wang, S., et al.: Twin‐field quantum key distribution over 830‐km fibre.
Inf. 3(1), 25 (2017). [Link] Nat. Photonics 16(2), 154–161 (2022). [Link]
322. Liu, Q., et al.: Advances in chip‐based quantum key distribution. En- s41566‐021‐00928‐2
tropy 24(10), 1334 (2022). [Link] 344. Sharma, P., et al.: Quantum key distribution secured optical networks: a
323. Chen, Y.‐A., et al.: An integrated space‐to‐ground quantum communi- survey. IEEE Open J. Commun. Soc. 2, 2049–2083 (2021). [Link]
cation network over 4,600 kilometres. Nature 589(7841), 214–219 org/10.1109/ojcoms.2021.3106659
(2021). [Link] 345. Moghaddam, E.E., Beyranvand, H., Salehi, J.A.: Resource allocation in
324. Vergoossen, T., et al.: Modelling of satellite constellations for trusted space division multiplexed elastic optical networks secured with quan-
node QKD networks. Acta Astronaut. 173, 164–171 (2020). https:// tum key distribution. IEEE J. Sel. Area. Commun. 39(9), 2688–2700
[Link]/10.1016/[Link].2020.02.010 (2021). [Link]
325. Van Meter, R., et al.: Path selection for quantum repeater networks. 346. Jia, J., et al.: Cost‐optimization‐based quantum key distribution over
Networking Sci. 3(1‐4), 82–95 (2013). [Link] quantum key pool optical networks. Entropy 25(4), 661 (2023). https://
013‐0026‐2 [Link]/10.3390/e25040661
326. Cirac, J.I., et al.: Quantum state transfer and entanglement distribu- 347. Niu, J., et al.: Noise‐suppressing channel allocation in dynamic DWDM‐
tion among distant nodes in a quantum network. Phys. Rev. Lett. QKD networks using lightgbm. Opt. Express 27(22), 31741–31756
78(16), 3221–3224 (1997). [Link] (2019). [Link]
3221 348. Niu, J., et al.: Key‐size‐driven wavelength resource sharing scheme for
327. Chanelière, T., et al.: Storage and retrieval of single photons transmitted QKD and the time‐varying data services. J. Lightwave Technol. 39(9),
between remote quantum memories. Nature 438(7069), 833–836 (2005). 2661–2672 (2021). [Link]
[Link] 349. Wang, R., et al.: AI‐enabled large‐scale entanglement distribution
328. Le, L., et al.: Entanglement routing for quantum networks: a deep quantum networks. In: Optical Fiber Communication Conference, pp.
reinforcement learning approach. In: ICC 2022‐IEEE International Tu1I–4. Optica Publishing Group (2021)
Conference on Communications (2022) 350. Nejabati, R., Wang, R., Simeonidou, D.: Dynamic quantum network:
329. Damgård, I.B., et al.: Secure identification and QKD in the bounded‐ from quantum data centre to quantum cloud computing. In: Optical
quantum‐storage model. In: Advances in Cryptology‐CRYPTO Fiber Communication Conference, pp. Th3D–1. Optica Publishing
2007: 27th Annual International Cryptology Conference, Santa Bar- Group (2022)
bara, CA, USA, August 19‐23, 2007. Proceedings 27, pp. 342–359. 351. Bahrani, S., Razavi, M., Salehi, J.A.: Wavelength assignment in hybrid
Springer (2007) quantum‐classical networks. Sci. Rep. 8(1), 3456 (2018). [Link]
330. Fitzsimons, J.F.: Private quantum computation: an introduction to blind org/10.1038/s41598‐018‐21418‐6
quantum computing and related protocols. npj Quantum Inf. 3(1), 23 352. Bahrani, S., et al.: Wavelength assignment in quantum access networks
(2017). [Link] with hybrid wireless‐fiber links. JOSA B 36(3), B99–B108 (2019).
331. Broadbent, A.: Delegating private quantum computations. Can. J. Phys. [Link]
93(9), 941–946 (2015). [Link] 353. Cao, Y., et al.: Reinforcement learning based multi‐tenant secret‐key
332. Chuang, I.L.: Quantum algorithm for distributed clock synchronization. assignment for quantum key distribution networks. In: 2019 Optical
Phys. Rev. Lett. 85(9), 2006–2009 (2000). [Link] Fiber Communications Conference and Exhibition (OFC), pp. 1–3.
physrevlett.85.2006 IEEE (2019)
333. Cuomo, D., Caleffi, M., Cacciapuoti, A.S.: Towards a distributed 354. Sparkes, B., et al.: Gradient echo memory in an ultra‐high optical depth
quantum computing ecosystem. IET Quantum Commun. 1(1), 3–8 cold atomic ensemble. New J. Phys. 15(8), 085027 (2013). [Link]
(2020). [Link] org/10.1088/1367‐2630/15/8/085027
334. Illiano, J., et al.: Quantum internet protocol stack: a comprehensive 355. Mottola, R., Buser, G., Treutlein, P.: Optical memory in a micro-
survey. Comput. Netw. 213, 109092 (2022). [Link] fabricated rubidium vapor cell. Phys. Rev. Lett. 131(26), 260801 (2023).
comnet.2022.109092 [Link]
 26328925, 2024, 3, Downloaded from [Link] by Pondicherry University, Wiley Online Library on [10/08/2025]. See the Terms and Conditions ([Link] on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
MAFU
- 231

356. Cho, Y.‐W., et al.: Highly efficient optical quantum memory with long 363. Khatri, S.: Policies for elementary links in a quantum network. Quantum
coherence time in cold atoms. Optica 3(1), 100–107 (2016). [Link] 5, 537 (2021). [Link]
org/10.1364/optica.3.000100 364. Yun‐Hong, G., et al.: Prediction and experimental verification for
357. Chen, Y.‐H., et al.: Coherent optical memory with high storage effi- satellite‐to‐ground quantum communication key rate based on machine
ciency and large fractional delay. Phys. Rev. Lett. 110(8), 083601 (2013). learning. J. Infrared Millim. Waves 40(3), 420–425 (2021)
[Link] 365. Robertson, E., et al.: Machine learning optimal control pulses in an
358. Reim, K., et al.: Single‐photon‐level quantum memory at room tem- optical quantum memory. In: European Quantum Electronics Confer-
perature. Phys. Rev. Lett. 107(5), 053603 (2011). [Link] ence, p. eb_10_3. Optica Publishing Group (2023)
1103/physrevlett.107.053603 366. Shinbrough, K., Lorenz, V.O.: Variance‐based sensitivity analysis of λ‐
359. Jobez, P., et al.: Coherent spin control at the quantum level in type quantum memory. Phys. Rev. 107(3), 033703 (2023). [Link]
an ensemble‐based optical memory. Phys. Rev. Lett. 114(23), org/10.1103/physreva.107.033703
230502 (2015). [Link] 367. I. S. G. (SG13): Future networks. (2023)
360. Sparkes, B., et al.: ac Stark gradient echo memory in cold atoms.
Phys. Rev. 82(4), 043847 (2010). [Link]
043847
361. Leung, A., et al.: Extending gradient echo memory using machine learning How to cite this article: Mafu, M.: Advances in
and single photons. In: Conference on Lasers and Electro‐Optics/Pacific
artificial intelligence and machine learning for quantum
Rim, pp. Th1D–2. Optica Publishing Group (2018)
362. Buchler, B., et al.: Stopped and stationary light with cold atomic en- communication applications. IET Quant. Comm. 5(3),
sembles and machine learning. In: CLEO: QELS_Fundamental Science, 202–231 (2024). [Link]
pp. FM1G–5. Optica Publishing Group (2018)