Peer-to-Peer Networking and Applications (2025) 18:307

[Link]

RESEARCH

A hybrid framework for secure IoT communication using lightweight

cryptography and machine learning-based authentication
Geeta Sharma1,4 · Narinder Verma3 · Sanjeev Jain1 · Radhe Shyam Sharma2

Received: 3 July 2025 / Accepted: 17 September 2025

© The Author(s), under exclusive licence to Springer Science+Business Media, LLC, part of Springer Nature 2025

Abstract
Securing data communication in Internet of Things (IoT) networks is challenging due to limited power, processing capacity,
and the heterogeneous nature of devices. Traditional encryption methods are often resource-intensive, while conventional
authentication schemes lack adaptability, limiting their applicability in mission-critical domains such as healthcare, smart
surveillance, and autonomous systems. This study presents a unified security framework that integrates lightweight Hybrid
Encryption Algorithms (HEA) with machine learning–based authentication to enhance secure communication across
diverse sensor nodes. A real-time testbed comprising temperature sensors, ECG modules, and Raspberry Pi-based image
sensors was developed for evaluation. The proposed hybrid encryption scheme, optimized for low-power, low-memory
IoT devices, outperforms AES and RSA in speed and efficiency, while a novel S-box design improves non-linearity and
cryptanalysis resistance. A machine learning-driven authentication module enables real-time anomaly detection and access
control, adapting dynamically to device behaviour patterns. Comparative analysis against benchmark algorithms, including
PRESENT, SIMON, SPECK, TEA, and a custom S-box-based approach, demonstrates lower latency, reduced computational
overhead, and improved scalability. The results validate the framework’s practicality for secure, low-latency communication
in next-generation IoT applications.

Keywords Hybrid encryption · Cryptography · Machine learning · Cybersecurity · IoT

1 Introduction

The rapid advancement of IoT-based robotics has revolutionized

numerous domains, including healthcare, manufacturing,
defence, and logistics [1]. IoT devices are increasingly being
deployed in critical applications such as surgical assistance,
Narinder Verma
[Link]@[Link] autonomous navigation, and industrial automation, where
they interact with sensitive data and perform tasks that directly
Geeta Sharma
geetaalam8@[Link] impact human safety and operational efficiency [2]. However,
this growing reliance on IoT systems has also made them
Sanjeev Jain
[Link]@[Link] attractive targets for cyberattacks. Malicious actors can exploit
vulnerabilities to steal sensitive data, disrupt operations, or
Radhe Shyam Sharma
radheshyam@[Link] even cause physical harm, underscoring the urgent need
for robust security mechanisms [3]. The heterogeneity and
1
Department of Computer Science and Information resource-constrained nature of IoT devices further increase the
Technology, Central University of Jammu, Rahya Suchani, complexity of securing these systems. Each device often has
Bagla, Jammu 181143, India
a different architecture and security capability. Sensor nodes
2
Indian Institute of Technology, Mandi, India are limited by low power, low memory, and low processing
3
School of Computer Science and Engineering, IILM capacity, making it challenging to implement traditional
University, Greater Noida 201306, India encryption techniques without compromising performance
4
School of Computer Science Engineering and Technology, [4–8]. To address these constraints, there is a pressing need to
Bennett University, 201310 Greater Noida, India

13
307 Page 2 of 18 Peer-to-Peer Networking and Applications (2025) 18:307

adopt lightweight encryption techniques that provide strong ● Proposed a machine learning-based node authentication
cryptography while maintaining the confidentiality, integrity, system that detects anomalies and unauthorized access
and privacy of data [9–12]. with high accuracy, dynamically learning from node
To address these challenges, this research proposes a behaviour patterns.
novel architecture that integrates lightweight encryption and ● Integrated the encryption and authentication mechanisms
machine learning-based authentication to secure IoT systems. into a unified framework capable of end-to-end protection
The proposed framework is built on three key aspects: — from data generation at IoT nodes to reception at the
cloud/server, addressing both confidentiality and trust.
(1) Sensor Input: where data is collected from distributed ● Demonstrated practical implementation on a real
IoT devices and sensors; network of heterogeneous IoT nodes in a client-side
(2) Data Encryption: where a lightweight hybrid robotics environment, validating real-world feasibility
encryption technique is employed to protect the data and performance.
during transmission; and. ● Evaluation of the proposed framework based on key security
(3) Authentication using Machine Learning: where and performance metrics, such as encryption efficiency,
intelligent models are used to verify the legitimacy of authentication accuracy, and computational overhead.
communicating nodes and detect anomalies in real time. ● Comparative analysis with existing encryption
techniques to highlight suitability for resource-
Let D = {d₁, d₂, …, dₙ} be the set of data packets generated constrained IoT environments.
by n IoT nodes, and N = {n₁, n₂, …, nₙ} be the corresponding ● Integration of IoT-based robotics security considerations,
set of IoT nodes. Each data packet di ∈ D is encrypted using focusing on critical applications like healthcare,
the proposed lightweight hybrid encryption algorithm Eₖ that autonomous navigation, and industrial automation,
uses a cryptographic key k and an S-box transformation S. The where secure data transmission is essential.
encryption function E maps each data packet to its encrypted
form C, such that C = E(D). To ensure that only legitimate The remainder of this paper is organized as follows:
nodes send data, an authentication function A: N → {0, 1} Sect. 2 reviews related work, highlighting the strengths
is utilized, which returns 1 if a node ni is authenticated and and limitations of existing approaches. Section 3 describes
0 otherwise. This function is implemented using a machine the proposed methodology, detailing the encryption and
learning model 𝓜 trained on features F(ni) extracted from authentication processes. Section 4 presents experimental
the behaviour of node ni. The objective of this framework is to results and comparative analysis, followed by conclusions and
design a secure, intelligent, and resource-efficient system that future directions in Sect. 5.
maximizes data confidentiality and authentication accuracy
while minimizing computational overhead, all within the
strict resource constraints of IoT nodes. This integrated 2 Related work
approach ensures both the trustworthiness of the data source
and the confidentiality of transmitted information, making it This section presents recent studies related to the hybrid
particularly well-suited for mission-critical IoT applications. encryption techniques. A hybrid encryption approach that
combines the Blowfish symmetric algorithm with Elliptic Curve
1.1 Contributions Cryptography (ECC) is presented in [13]. This method aims
to secure data transmission in IoT devices while minimizing
The following are the main contributions of the study: computational overhead. The results demonstrated that this
combination provides a balanced trade-off between security
● Introduced a custom hybrid encryption scheme and performance, making it suitable for resource-limited
specifically optimized for low-power and low-memory IoT devices. Another study proposed a lightweight hybrid
IoT nodes, outperforming standard AES and RSA cryptographic framework utilizing the Advanced Encryption
combinations in terms of speed and resource usage. Standard (AES) in conjunction with ECC. This framework
Development of a novel framework that integrates addresses the need for secure communication in IoT-enabled
hybrid encryption and machine learning-based systems by ensuring data confidentiality and integrity without
authentication to enhance IoT-based robotics security. imposing significant computational burdens on the devices [14].
● Developed a novel S-box structure tailored to the Authentication remains a critical aspect of IoT security, ensuring
proposed encryption algorithm to improve non-linearity that only legitimate devices can access the network. Machine
and resistance against linear and differential cryptanalysis, learning (ML) techniques have been employed to enhance
significantly increasing cryptographic strength. authentication processes by analysing patterns and behaviours

13
Peer-to-Peer Networking and Applications (2025) 18:307 Page 3 of 18 307

associated with device interactions. A comprehensive survey devices with different processing capabilities and security
highlighted the application of reinforcement learning methods, requirements. Many encryption algorithms are designed for
such as Q-learning, for IoT device authentication. These methods different IoT architectures, but they fail to account for the
detect anomalies and potential security breaches by learning varying computational capacities of smart sensors and industrial
from the environment, thereby enhancing the robustness of the IoT devices. In the context of machine learning for secure
authentication process [15]. authentication, several gaps exist in terms of model reliability,
Further, a federated machine learning approach was adaptability, and training data quality. Many authentication
developed to bolster authentication and authorization in IoT models depend on large, labeled datasets, but obtaining high-
ecosystems. This system utilizes data collected from IoT quality IoT authentication datasets remains a challenge due
devices to train models that can identify and authenticate to privacy concerns and data variability. Additionally, most
devices based on their unique physical layer characteristics, machine learning-based authentication frameworks assume
thereby enhancing security without compromising user privacy stationary threat environments, making them ineffective against
[16, 17]. Table 1 presents an analysis of existing studies. emerging zero-day attacks and adversarial manipulations.
Despite significant advancements in securing IoT-based
systems through various encryption techniques, several
challenges still persist. One of the primary challenges is the 3 Proposed approach
computational overhead introduced by hybrid encryption
schemes. While combining symmetric and asymmetric The proposed system architecture comprises two primary
encryption enhances security, it often results in increased modules: the client side and the server side. At the client side,
processing time and energy consumption, which is problematic data is generated by various sensors and transmitted to the
for resource-constrained IoT devices. Additionally, existing server side for storage. However, during transmission, the data
research primarily focuses on optimizing encryption techniques is vulnerable to potential intrusions. To safeguard the data from
without fully considering real-time constraints, making it unauthorized access, encryption is applied before transmission.
difficult to implement these methods in real-time IoT networks. Prior to storage on the server side, the system verifies whether
Furthermore, most studies evaluate encryption schemes using the data originates from an authenticated node or not. The
traditional security metrics but fail to assess their performance proposed framework implements an adaptive hybrid Encryption
under real-world adversarial conditions or large-scale and intelligent authentication framework for heterogeneous
deployments. Device heterogeneity in IoT networks presents networks, ensuring both data confidentiality and authenticity.
another critical challenge. IoT ecosystems consist of diverse The basic flow of the framework is shown in Fig. 1. Data from

Table 1 Summary of existing studies

Ref Encryption tech- Application Device Description Performance Limitations
nique used Context Employed
[18] ECC-based healthcare Simulation Encrypts healthcare robotics data using Improved key generation, Dependency
encryption, ECC for security and HMAC-SHA1 for encryption/decryption time, on trusted third
HMAC-SHA1 for data integrity, ensuring efficient cloud- reduced security overhead, parties, internal
integrity robot communication. suitable for low-power devices. threat mitigation
needs further
improvement.
[19] Lightweight Industrial ARM Cortex-M4 Combines AES for data encryption and Encryption speed: 12 Mbps, High memory
Hybrid AES-ECC Robotic Arms Microcontroller ECC for key exchange, optimized for Energy consumption: 0.5 mJ usage for ECC key
real-time systems. per operation. storage, limited
scalability.
[20] Hybrid encryp- Healthcare Simulation Hybrid encryption is used for security, Overall execution time = Computational
tion (ECC, AES, and elliptic curve digital signature for 134.96 ms overhead on each
Serpent) data integrity. entity; lacks fine-
grained access
control.
[21] Improved ECC, Healthcare Simulation Combines biometric authentication with Encryption time: 1.032 µs, Lacks real-time
SHA-512, SHA-512 for integrity, uses improved Decryption time: 1.004 µs, implementation.
Substitution- ECC with an additional secret key for Correlation coefficient: 0.045,
Caesar cipher enhanced encryption security. Lower computational cost than
RSA and ECC.
[22] Lightweight Medical Simulation Encrypts medical images using an opti- Security level: 97.82%, Limited to the
Hybrid mized S-box block cipher and elliptic encryption time: 15.4s, brain MRI image
encryption curves, then conceals them in a cover execution time: 18.41s, energy only
image using LSB before cloud storage. consumption: 1.33 pJ/bit

13
307 Page 4 of 18 Peer-to-Peer Networking and Applications (2025) 18:307

Fig. 1 Flow diagram for the proposed framework

heterogeneous IoT devices is encrypted at the client side using 3.2 Server-side
the proposed HEA algorithm, while machine learning–based
authentication verifies node legitimacy. Authenticated data is In the proposed framework, data transmitted from the client
decrypted at the server and then securely stored or used for side to the server side is transferred in the form of packets
actuation, ensuring end-to-end security in resource-constrained and frames, which contain information such as frame headers,
IoT environments. Each module employed in the network frame footers, IP headers, and data payload. This frame
architecture is elaborated in the subsequent subsections. information is divided into two parts: the data block and the
auxiliary information block. The system extracts the data
3.1 Client-side block separately while isolating the auxiliary information
block to verify the node’s authentication status.
The client side utilizes various sensors integrated with distinct The server accepts incoming data only if the transmitting
processors to create a heterogeneous environment. The client- node is authenticated by the pre-trained model. If the
side process is further subdivided into two sub-categories, as data is received from an unauthenticated node, the server
depicted in Fig. 2. The first phase involves data Capturing, discards the frame to prevent unauthorized access. The
where three distinct sensor nodes are employed to establish pre-trained model checks the node’s authentication status
a real-time heterogeneous environment. These nodes include based on the auxiliary information block. If the node is
a camera sensor for capturing image data, a temperature and authenticated, the server receives the encrypted data,
humidity sensor for generating digital data, and an ECG sensor applies the decryption algorithm, and stores the original
for producing analog signals. This diverse data collection data securely on the server side, as illustrated in Fig. 3.
facilitates the simulation of a heterogeneous IoT network by
providing different data types. The second phase is the data 3.3 Hybrid encryption algorithm HEA
encryption phase, where the captured sensor data is encrypted
using the proposed novel lightweight Hybrid Encryption In this subsection, the novel algorithm employed at the client
Algorithm (HEA). The encryption process converts the sensor side to provide data security is explained. The proposed
data into a secure coded format to ensure data confidentiality encryption algorithm is designed to operate on 64-bit input
during transmission. The HEA is specifically designed to data, as depicted in Fig. 4. Initially, the input data received
accommodate the constrained computational resources of IoT from the sensors is converted into its binary representation
devices, making it more suitable than conventional encryption of 64 bits, where each bit is labelled as B0, B1, …, B63. The
algorithms. Implementing encryption at the physical layer 64-bit data is then divided into sixteen 4-bit blocks, denoted as
(node level) enhances the security of data by preventing B0, B1, …, B15. In the first substitution layer, each 4-bit block
unauthorized access and mitigating various attacks that may undergoes substitution using the novel S-Box 1, producing the
occur during data transmission. substituted blocks labelled as Bs0, Bs1, …, Bs15. The second
This approach significantly improves the overall substitution layer is applied to the output of the first layer,
resilience of the IoT network without imposing excessive where each block is substituted again using the novel S-Box
computational overhead on resource-constrained devices. 2, generating the updated 64-bit data. Subsequently, an XOR

13
Peer-to-Peer Networking and Applications (2025) 18:307 Page 5 of 18 307

Fig. 2 Framework for data security using novel lightweight encryption algorithm HEA at client-side

13
307 Page 6 of 18 Peer-to-Peer Networking and Applications (2025) 18:307

operation is performed between the 64-bit data and a 64-bit the round key. To generate keys for the subsequent 19 rounds,
round key to introduce randomness. This process is iterated a 16-bit right circular shift operation is applied to the 128-bit
for 20 rounds to enhance the randomness and unpredictability key. After each shift, the right-most 64 bits are selected as the
of the ciphertext. After completing all rounds, the final output round key. This process is repeated iteratively to produce a total
represents the encrypted 64-bit data, ensuring high security of 20 round keys, ensuring a unique key for each encryption
and resistance against cryptographic attacks. In Algorithm 1, a round, thereby enhancing the security and randomness of the
step-by-step procedure to encrypt the data is explained. For the algorithm. A graphical representation of the same procedure is
first round, the right-most 64 bits are extracted and utilized as shown in Fig. 5.

Algorithm 1. Step by step procedure to encrypt data using HEA

Input: 64-bit block, and 64-bit register key
Output: 64-bit encrypted block
1. Begin
2. Initialization:
inputblock = b63…..b1b0 // 64-bits input
registerkey = k64…..k1k0 // 80-bits key value
key= generateroundkey(registerkey) // update key value for each round

3. Encryption_function:
for i=1 to 20

a. Divide 64-bit data into 4-bit blocks: B0, B1, ..., B15
b. Perform substitution on each 4-bit block using S-Box 1
c. Perform substitution on each 4-bit block using S-Box 2
d. XOR the resulting 64-bit data with the round key
e. Return 64-bit message

4. End
The key generation process begins by selecting 128 ran-
dom bits. In algorithm 2, step by step procedure to generate
20 different key values is explained.

Algorithm 2. Key generation procedure for HEA

Input: 128 random bits
Output: 64-bit key
1. Begin
2. Initialization:
initial_key = k127,k126,…..k1k0 //128 bit value
round_keys = [ ] // empty list to store 20 rounds key-value
3. generate_round_keys(initial_key):
for first round:
Extract the right-most 64 bits of k and pass to round_keys list.
for i=2 to 20
a. Perform a 16-bit right circular shift on the 128-bits key k.
b. Extract the right-most 64 bits of the shifted key k.
c. Append 64-bits key value to round_keys list.
d. Return round_keys list.

4. End

13
Peer-to-Peer Networking and Applications (2025) 18:307 Page 7 of 18 307

Fig. 3 Framework for hecking authentication of nodes using intelligent approach at server-side

13
 307 Page 8 of 18 Peer-to-Peer Networking and Applications (2025) 18:307

Fig. 4 Proposed framework of

lightweight encryption algorithm
HEA

The Substitution Box (S-Box) plays a critical role in i. Non-Linearity and Confusion
lightweight encryption algorithms by introducing non- The primary function of the S-Box is to introduce
linearity into the cryptographic process, which is essential confusion into the encryption process by transforming
for enhancing the security and robustness of the encryption input bits into output bits in a nonlinear manner.
system. In lightweight encryption algorithms, particularly This transformation makes it difficult for attackers
those designed for resource-constrained devices such as IoT to predict the relationship between plaintext and
sensors, S-Box is carefully designed to provide high security ciphertext, thereby strengthening the encryption.
with minimal computational complexity. ii. Resistance Against Cryptographic Attacks

13
Peer-to-Peer Networking and Applications (2025) 18:307 Page 9 of 18 307

Fig. 5 Key-generation procedure

for the novel algorithm HEA

A well-designed S-Box improves the algorithm’s In Figs. 6 and 7, the total number of gate equivalents is
resistance to common cryptographic attacks such as described to implement the proposed s-box on a hardware
differential cryptanalysis and linear cryptanalysis. By platform.
providing non-linear mappings, the S-Box reduces the In Table 4, a comparative analysis for GE is performed
likelihood that small changes in the input will produce based on a number of different gate requirements. From
predictable changes in the output. Table 4, it can be inferred that the proposed s-boxes 1 and 2
iii. Randomness and Entropy utilized a smaller number of total GE as compared to other
The S-Box enhances the randomness of the ciphertext by existing standardized s-boxes.
creating highly unpredictable outputs. This randomness Evaluation of S-boxes based on different properties,
increases the entropy of the encrypted data, making it such as whether the S-box is perfectly nonlinear,
more challenging for attackers to reverse-engineer the balanced, permutable, etc., has been performed as shown
encryption process. in Fig. 8.
iv. Lightweight Design Comparative analysis of proposed S-boxes based on
In lightweight encryption algorithms, S-Boxes are often several properties with existing S-boxes is shown in Table 5.
designed to be small in size with simple lookup tables to From Table 5, it can be revealed easily that the proposed
minimize memory usage and computational overhead. S-boxes satisfied all the properties same as the benchmark
This makes the algorithm suitable for devices with S-boxes.
limited processing power and storage capacity, such as
IoT sensors and embedded systems. 3.4 Authentication using ML
v. Multiple Layers of Security
Lightweight encryption algorithms often employ The ML-based authentication system is designed to
multiple S-Box layers to increase the complexity of detect unauthorized or anomalous devices by leveraging
the encryption process without significantly affecting device-specific and communication-level features that
performance. This layered approach provides better serve as unique fingerprints. The dataset is compiled from
diffusion, ensuring that small changes in the plaintext real-time sensor streams originating from heterogeneous
result in significant changes in the ciphertext. sources such as Arduino-connected temperature/humidity
In the proposed encryption framework, two novel S-Boxes sensors, Raspberry Pi-based image sensors, and ECG
are utilized, each with distinct substitution patterns as modules. Figure 9, depicts the overall process. Feature
shown in Tables 2 and 3. This dual substitution process engineering involved the extraction of temporal, statistical,
enhances the algorithm’s security while maintaining and hardware-bound characteristics, including packet
low computational complexity, making it ideal for inter-arrival times, variance in analog sensor readouts,
heterogeneous networks with limited resources. transmission frequencies, unique hardware IDs, and noise

13
 307

13
Page 10 of 18

Table 2 Proposed novel S-box 1

x 0 1 2 3 4 5 6 7 8 9 A B C D E F
S[x] F 0 C 5 B 1 7 3 9 D 6 2 8 4 A E

Table 3 Proposed novel S-box 2

X 0 1 2 3 4 5 6 7 8 9 A B C D E F
S[x] 9 F D 5 B 2 E 8 4 6 C 0 3 A 1 7
Peer-to-Peer Networking and Applications
(2025) 18:307
Peer-to-Peer Networking and Applications (2025) 18:307 Page 11 of 18 307

Fig. 6 GE needs to implement s-box1

Fig. 7 GE needs to implement s-box2

signatures in analog-to-digital conversion. These features which offered resilience against minor fluctuations in fea-
are selected based on their non-replicable nature, which ture distributions. The RF model, built with 100 estimators
helps in establishing device identity with high fidelity and Gini impurity as the splitting criterion, provided robust
[23–25]. ensemble learning with improved generalization. All mod-
Three supervised classifiers, Decision Tree (DT), els are trained on 80% of the dataset and validated on the
K-Nearest Neighbors (KNN), and Random Forest (RF), are remaining 20%, using 10-fold cross-validation to prevent
employed to model the authentication layer. The DT classi- overfitting.
fier provided a baseline interpretability with its binary split- Quantitatively, the RF model outperformed others with
based decision rules. For KNN, Euclidean distance is used an authentication accuracy of 97.6%, a false acceptance
as the metric, and optimal performance is observed at k = 5, rate (FAR) of 1.8%, and a false rejection rate (FRR) of

13
 307 Page 12 of 18 Peer-to-Peer Networking and Applications (2025) 18:307

Table 4 A comparative analysis of various S-boxes based on gate 4 Performance evaluation

equivalents (GE)
S. Algorithm name Input Number Number Number
No. for S-box Size of NOT of NAND of NOR This section presents the performance evaluation of the
gates gates used gates proposed novel hybrid lightweight cryptography algorithm.
used used The objective is to examine the robustness, efficiency, and
1. Present 4 08 23 13 suitability of the algorithm in real-time environments with
2. Gift 4 08 19 16 constrained and heterogeneous devices.
3. Klein 4 06 21 16
4. Elephant 4 04 24 16
4.1 Experimental setup
5. Spongent 4 04 24 16
6. Proposed S-box 1 4 04 24 04
7. Proposed S-box 2 4 02 20 07
To evaluate the proposed algorithm, a real-time environment
was established using a range of microcontrollers and
sensors, as detailed in Table 6. The devices were chosen to
1.4%. KNN followed with 95.2% accuracy and DT with simulate a heterogeneous IoT environment, ensuring that the
93.8%. Average authentication latency across models is algorithm could operate efficiently across diverse hardware
< 6ms, confirming real-time suitability even on low-power architectures. The selected components included resource-
microcontrollers. constrained microcontrollers such as ESP32, Arduino, etc.,
This authentication layer operates in tandem with the integrated with various sensors, including temperature and
encryption modules (e.g., TEA, PRESENT, SIMON, humidity sensors, ECG sensors, etc., reflecting real-world
SPECK), such that encryption keys are only issued to IoT scenarios. The experimental setup incorporated several
devices classified as legitimate by the ML model. For key components to evaluate the performance of the proposed
instance, in the PRESENT algorithm, a device must pass lightweight cryptographic algorithm. Microcontrollers were
ML-based authentication before being permitted to execute deployed to assess the algorithm’s lightweight nature and
key-scheduling and S-box transformations. This mechanism computational feasibility on resource-constrained devices.
effectively filters out adversaries before cryptographic These microcontrollers were interfaced with various sensors
routines begin, mitigating the risk of key exposure or to simulate real-time data generation and transmission.
algorithm misuse. Additionally, a high-performance GPU-enabled server

Fig. 8 Different satisfied proper-

ties of s-boxes utilized in HEA
s-boxes

Table 5 A comparative analysis of various s-boxes based on different properties

S. Algorithm name for S-box Input Size Almost perfect nonlinear Balanced Linear-structure Per-mutable Involution
No. (APN)
1. Present 4 × √ √ √ ×
2. Gift 4 × √ √ √ ×
3. Klein 4 × √ × √ √
4. Skinny 4 × √ √ √ ×
5. Elephant 4 × √ √ √ ×
6. Spongent 4 × √ √ √ ×
7. Proposed S-box 1 4 × √ √ √ ×
8. Proposed S-box 2 4 × √ √ √ ×

13
Peer-to-Peer Networking and Applications (2025) 18:307 Page 13 of 18 307

Fig. 9 A general architecture for ML authentication process

system was utilized for centralized processing, including e) Security Metrics: Analysis confirmed resistance
authentication checking of nodes, decryption of the data, to common lightweight cryptanalysis techniques,
and storage of information. The server was configured with ensuring that the hybrid approach maintains security
39 M Cache, 12 × 16 GB RDIMM Dual Rank RAM, and without compromising speed or resource efficiency.
4 × 600 GB 15 K SAS HDD. Table 5 shows the detailed
description of the devices utilized to create a heterogeneous The overall results affirm that the proposed hybrid lightweight
real-time network. cryptography algorithm is practical for deployment in real-
world IoT scenarios, offering a promising balance between
4.2 Experimental results security, performance, and resource optimization. The
summarized results are presented in Table 7; Fig. 10. Further,
The proposed cryptographic algorithm was implemented among the three evaluated classification algorithms, RF
and deployed across the environment. The results highlight achieved the best performance with over 98% accuracy,
the algorithm’s efficiency, lightweight processing footprint, indicating high reliability and balance.
and robustness against computational limitations of To address the trade-offs associated with the proposed
constrained devices. HEA, we have incorporated a detailed analysis highlighting
Key observations include: the balance between security and computational efficiency,
particularly in the context of resource-constrained IoT
a) Execution Time: The encryption and decryption operations environments. While our custom S-box-based encryption
demonstrated low latency across microcontrollers, achieves notably low encryption latency (as low as 3.6
remaining well within acceptable thresholds for real-time nanoseconds), we acknowledge that this efficiency may
applications. involve compromises in cryptographic strength when
b) Memory Consumption: The algorithm exhibited compared to more robust, yet resource-intensive, standards.
minimal RAM and flash usage, validating its The HEA is therefore designed to ensure sufficient data
suitability for low-power and memory-constrained protection while remaining lightweight enough for real-
devices. time applications such as wearable healthcare devices,
c) Energy Efficiency: Power profiling of the microcon- autonomous systems, and smart industrial platforms.
trollers showed a negligible increase in energy consump- Additionally, we have elaborated on the challenges
tion, indicating optimized performance. associated with training machine learning models for
d) Cross-Platform Compatibility: The algorithm func- authentication in IoT scenarios, specifically the issues of
tioned consistently across all heterogeneous devices, limited, imbalanced, and heterogeneous datasets. To address
confirming its robustness and flexibility. these, we incorporate data augmentation techniques and

13
 307 Page 14 of 18 Peer-to-Peer Networking and Applications (2025) 18:307

Table 6 Detailed description of the real-time environment devices

Component Model/Type Specification/Details Power Source Image

Microcontroller Arduino Uno ATmega328P, 5V logic, 14 Digital I/O, 6 USB (5V)

Board R3 Analog In

Temp/Humidity DHT11 Temp: 0–50°C, Humidity: 20–90% RH, Powered via

Sensor Digital Output Arduino 5V

ECG Sensor AD8232 Heart rate monitoring, Analog output, 3- Powered via
Module lead ECG ESP 8266
module

Robotic Platform AlphaBot2 N20 micro gear motor, 6V/600RPM, 5V USB

42×19mm rubber wheels, sensors, RGB charging
LEDs, battery status

Camera Pi Camera 8 MP (v2) / 12 MP (v3), video and still Powered via

Module v2/v3 capture Raspberry Pi

Microcontroller ESP 802.11 b/g/n Supports STA/AP/STA+AP External 5V

Board 8266MOD modes, WPA/WPA2 security (USB or VIN)

Server Dell EMC Processor with 39M Cache, 12× 16 GB Dedicated

PowerEdge RDIMM Dual Rank RAM, 4× 600 GB server power
R750 15K SAS HDD supply

online learning mechanisms that enable the models to adapt robustness of the proposed framework in ensuring secure and
over time and generalize across varying device behaviours. reliable communication in next-generation IoT environments.
Furthermore, we recognize the increasing risk of adversarial
attacks in IoT networks. To strengthen our system’s resilience,
the authentication framework has been extended to include 5 Comparative analysis
anomaly detection and adversarial training techniques,
enabling the model to better detect spoofing, poisoning, This section provides a comparative analysis of existing
and other manipulation attempts. These enhancements studies with the proposed approach, highlighting their
collectively reinforce the practicality, adaptability, and strengths and limitations.

13
Peer-to-Peer Networking and Applications (2025) 18:307 Page 15 of 18 307

Table 7 Key observations of the proposed hybrid encryption algorithm models that merge the strengths of symmetric and asymmetric
Device Data Encryption Encryption Decryp- File techniques. In [13], a novel encryption framework was presented,
Type Type Technique Time (avg, tion Time Size
ns) (avg, ns) (KB/ integrating Blowfish with elliptic curve cryptography (ECC).
MB) This combination leverages the speed of symmetric encryption
Temp Sen- Digital PRESENT 207.2 ns ~ 210 ns ~ 0.1 and the robust security of asymmetric methods, ultimately
sor + Ardu- (Sen- KB enhancing system throughput and reducing processing delays.
ino sor
Similarly, the authors in [26] developed a hybrid cryptographic
Data)
Temp Sen- Digital TEA 264.8 ns ~ 265 ns ~ 0.1
protocol that employs AES alongside ECC to facilitate secure
sor + Ardu- (Sen- KB data sharing. Their method also incorporates a Message
ino sor Authentication Code (MAC) to validate data integrity and
Data) authenticity. The encrypted data is stored on the cloud, helping
Temp Sen- Digital SIMON 166.4 ns ~ 170 ns ~ 0.1 to optimize storage efficiency. Experimental results validate the
sor + Ardu- (Sen- KB
ino sor proposed approach’s effectiveness, outperforming traditional
Data) solutions in various performance metrics. Furthermore, the
Temp Sen- Digital SPECK 165.2 ns ~ 170 ns ~ 0.1 study in [27] focuses on improving IoT security by utilizing a
sor + Ardu- (Sen- KB hybrid model that fuses the Advanced ElGamal signature with
ino sor
the Data Encryption Standard (DES).
Data)
Temp Sen- Digital Custom 3.6 ns ~ 4 ns ~ 0.1 The encryption process is designed to be non-
sor + Ardu- (Sen- SBOX KB deterministic, producing unique ciphertexts even when
ino sor encrypting identical plaintext multiple times, thanks to its
Data) dynamic round key generation. The implementation, carried
Raspberry Image PRESENT ~ 800,000 ~ 810,000 ~ 500
out using Python, reported strong performance indicators:
Pi + Camera (JPEG/ ns (0.8 ms) ns (0.81 KB
PNG) ms) – 2 MB a throughput of 99.68%, a 100% transmission rate, and
Raspberry Image TEA ~ 1,000,000 ~ 1 ms ~ 500 a delay time of just 0.3 s. Authors in [28] developed a
Pi + Camera (JPEG/ ns (1 ms) KB hybrid encryption model for IoT-based irrigation systems
PNG) – 2 MB utilizing ECC, RC4, and SHA-256 to protect sensitive data.
Raspberry Image SIMON ~ 700,000 ~ 0.7 ms ~ 500 Simulation results indicate that the proposed model reduces
Pi + Camera (JPEG/ ns (0.7 ms) KB
PNG) – 2 MB encryption and decryption time compared to existing studies.
Raspberry Image SPECK ~ 680,000 ~ 0.68 ms ~ 500 Table 8 depicts a comparative analysis with existing studies.
Pi + Camera (JPEG/ ns (0.68 KB
PNG) ms) – 2 MB
ECG Sen- Signal PRESENT ~ 350 ns ~ 360 ns ~ 1 KB 6 Conclusion
sor + Micro- (ECG (per
controller wave- read-
form) ing) In the expanding domain of IoT and embedded systems,
ECG Sen- Signal TEA ~ 400 ns ~ 410 ns ~ 1 KB ensuring secure and timely data transmission remains a
sor + Micro- (ECG (per pressing challenge, particularly in environments constrained
controller wave- read- by limited processing power and energy resources. This study
form) ing)
explored the effectiveness of lightweight encryption methods
ECG Sen- Signal SIMON ~ 320 ns ~ 330 ns ~ 1 KB
sor + Micro- (ECG (per and machine learning-based authentication to address these
controller wave- read- limitations. A detailed evaluation was carried out on embedded
form) ing) platforms using real-time sensor data from temperature, ECG,
ECG Sen- Signal SPECK ~ 310 ns ~ 320 ns ~ 1 KB and image-based modules. Among the encryption algorithms
sor + Micro- (ECG (per
controller wave- read- assessed PRESENT, TEA, SIMON, SPECK, and a custom-
form) ing) designed S-box-based method the proposed method was
evaluated and found to outperform the others in terms of
The proposed approach, which integrates encryption with encryption speed, achieving a minimum encryption time of 3.6
machine learning, addresses key challenges such as throughput, nanoseconds. This result highlights its suitability for real-time
latency, and scalability while maintaining very high security applications, though it may involve trade-offs in cryptographic
levels. Throughput is compared based on different data sizes robustness. To reinforce data integrity and access control,
(50 kb, 1024 kb, 2048 kb) and is optimized. The decryption machine learning models were integrated into the system to
process is similar to encryption but uses the subkeys in reverse perform user authentication and anomaly detection. These
order. Several recent studies have explored hybrid encryption models achieved over 96.5% classification accuracy, enhancing

13
 307 Page 16 of 18 Peer-to-Peer Networking and Applications (2025) 18:307

Fig. 10 Machine learning results

for intrusion detection

Table 8 Comparative analysis with existing studies

Ref Technique Throughput/Performance Data type Application Environment Client Evaluation metrics Combi-
supported server nation
with ML
[29] Chaotic maps, Low Image Healthcare Personal ⨯ Entropy, Correlation, ✔
Discrete DWT, Computer Chi-square analysis,
S-box, and Energy, Computational
DRPE complexity
[13] Hybrid Encryp- High General IoT Simulation ⨯ Execution Time, ⨯
tion (Blowfish + IoT data Ecosystems Throughput, Security
ECC) Level
[26] Hybrid Moderate IoT Com- IoT Cloud ✔ Data Security, Perfor- ⨯
Approach (AES munication Networks mance Efficiency
+ ECC) Data
[27] Lightweight 99.68% IoT Device Smart IoT Personal ✔ computational time, ⨯
hybrid (AES + Data systems Computer encryption and decryp-
DES) tion time,, transmission
rat
[28] ECC, RC4, High Agriculture Smart Simulation ✔ Encryption, decryption ⨯
and SHA-256 Sensors Irrigation time, and throughput
System
Pro- HEA Low Heter- Smart IoT Real time ✔ Encryption, decryption ✔
posed ogenous system time
work Sensors

the reliability of the system against unauthorized access. Acknowledgments The authors gratefully acknowledge the support
and facilities provided by the Department of Computer Science
The combination of lightweight cryptography and adaptive
and Information Technology, Central University of Jammu, Rahya-
ML-based security yielded a practical, low-latency solution Suchani (Bagla), Jammu – 181143, India, which greatly contributed to
for data protection in resource-constrained environments. the completion of this research.
The results confirm that the proposed framework can support
secure, real-time communication in critical applications such Author contributions G.S. was involved in data collection, initial
drafting, and manuscript review. N.V. carried out the formal analysis,
as wearable health devices, environmental monitoring, and validation, visualization, and contributed to writing and editing. S.J.
smart industrial systems. Looking ahead, the development of and R.S.S. provided overall supervision and guidance.
context-aware encryption strategies and the deployment of
edge-level intelligence will further strengthen the scalability Funding Not applicable.
and adaptability of security mechanisms in next-generation
IoT ecosystems. Data availability No datasets were generated or analysed during the
current study.

13
Peer-to-Peer Networking and Applications (2025) 18:307 Page 17 of 18 307

Declarations taxonomy, challenges and future research direction. Sensors

21(15):5122. ​h​t​t​p​​s​:​/​​/​d​o​i​​.​o​​r​g​/​​1​0​.​3​​3​9​0​​/​s​2​​1​1​5​5​1​2​2
16. Ahmed KI, Tahir M, Lau SL, Habaebi MH, Ahad A, Mughees
Competing interests The authors declare no competing interests.
A (2024) Trust-aware authentication and authorization for iot:
a federated machine learning approach. IEEE Internet Things J
12(8):9889–9904. ​h​t​t​p​​s​:​/​​/​d​o​i​​.​o​​r​g​/​​1​0​.​1​​1​0​9​​/​J​I​​O​T​.​2​0​2​4​.​3​5​1​2​6​5​7
References 17. Kumar M et al (Dec. 2023) A smart privacy preserving
framework for industrial IoT using hybrid meta-heuristic
algorithm. Sci Rep 13(1):1–17. ​h​t​t​p​​s​:​/​​/​d​o​i​​.​o​​r​g​/​​1​0​.​1​​0​3​8​​/​S​4​​1​5​9​
1. Pasdar A, Koroniotis N, Keshk M, Moustafa N, Tari Z (2024)
8​-​0​2​3​-​3​2​0​9​8​-​​2​​​​​;​S​U​B​J​M​E​T​A​= ​1​0​4​2​,​1​1​7​,​2​5​8​,​6​3​9​,​7​0​5​;​K​W​R​D​= ​
Cybersecurity solutions and techniques for internet of things
C​O​M​P​U​T​A​T​I​O​N​A​L​+​S​C​I​E​N​C​E​,​C​O​M​P​U​T​E​R​+​S​C​I​E​N​C​E​,​I​N​
integration in combat systems. IEEE Trans Sustain Comput. ​h​t​t​p​​s​:​/​​
F​O​R​M​A​T​I​O​N​+​T​E​C​H​N​O​L​O​G​Y​​​
/​d​o​i​​.​o​​r​g​/​​1​0​.​1​​1​0​9​​/​T​S​​U​S​C​.​2​0​2​4​.​3​4​4​3​2​5​6
18. Jain S, Doriya R (2022) Security framework to healthcare robots
2. Inuwa MM, Das R (2024) A comparative analysis of various
for secure sharing of healthcare data from cloud. Int J Inf Technol
machine learning methods for anomaly detection in cyber attacks
14(5):2429–2439. ​h​t​t​p​​s​:​/​​/​d​o​i​​.​o​​r​g​/​​1​0​.​1​​0​0​7​​/​S​4​​1​8​7​​0​-​0​​2​2​-​0​​0​9​​9​7​-​8​/​F​
on IoT networks. Internet Things 26:101162. ​h​t​t​p​​s​:​/​​/​d​o​i​​.​o​​r​g​/​​1​0​.​1​​
I​G​U​R​E​S​/​5
0​1​6​​/​J​.​​I​O​T​.​2​0​2​4​.​1​0​1​1​6​2
19. Fu B, Fang T, Zhang L, Zhou Y, Xiao H (2024) Communication
3. Md Rashed Buiya MS, Nuruzzaman Laskar AKM, Islam
security of intelligent information service platform combining
MR, Shil SK and R. E. R. S. and, Rahman MS, Chowdhury
AES and ECC algorithms. J Cyber Secur Technol 1–18. ​h​t​t​p​​s​:​/​​/​d​
(2021) Machine Learning in Business Analytics: Advancing
o​i​​.​o​​r​g​/​​1​0​.​1​​0​8​0​​/​2​3​​7​4​2​​9​1​7​​.​2​0​2​​4​.​​2​3​7​1​0​5​3
Statistical Methods for Data- Driven Innovation, no. pp. 104–
20. Das S, Namasudra S (2022) A novel hybrid encryption method to
111, 2023. ​h​t​t​p​​s​:​/​​/​d​o​i​​.​o​​r​g​/​​1​0​.​3​​2​9​9​​6​/​j​​c​s​t​s
secure healthcare data in IoT-enabled healthcare infrastructure.
4. Ali S, Anwer F (2024) Secure IoT framework for authentication
Comput Electr Eng 101:107991. ​h​t​t​p​​s​:​/​​/​d​o​i​​.​o​​r​g​/​​1​0​.​1​​0​1​6​​/​J​.​​C​O​M​​
and confidentiality using hybrid cryptographic schemes, Int. J.
P​E​L​​E​C​E​N​​G​.​​2​0​2​2​.​1​0​7​9​9​1
Inf. Technol., vol. 16, no. 4, pp. 2053–2067, Apr. ​h​t​t​p​​s​:​/​​/​d​o​i​​.​o​​r​g​/​​1​
21. Khan MA, Quasim MT, Alghamdi NS, Khan MY (2020) A secure
0​.​1​​0​0​7​​/​S​4​​1​8​7​​0​-​0​​2​4​-​0​​1​7​​5​3​-​W​/​F​I​G​U​R​E​S​/​1​4
framework for authentication and encryption using improved
5. Liao H et al (2024) A survey of deep learning technologies for
ECC for IoT-based medical sensor data. IEEE Access 8:52018–
intrusion detection in internet of things. IEEE Access 12:4745–
52027. ​h​t​t​p​​s​:​/​​/​d​o​i​​.​o​​r​g​/​​1​0​.​1​​1​0​9​​/​A​C​​C​E​S​S​.​2​0​2​0​.​2​9​8​0​7​3​9
4761. ​h​t​t​p​​s​:​/​​/​d​o​i​​.​o​​r​g​/​​1​0​.​1​​1​0​9​​/​A​C​​C​E​S​S​.​2​0​2​3​.​3​3​4​9​2​8​7
22. Dev BPV, Venkata Prasad K (2023) An adaptive lightweight
6. Khan AA et al (2025) Bdlt-iomt—a novel architecture: SVM
hybrid encryption scheme for securing the healthcare data
machine learning for robust and secure data processing in internet
in cloud-assisted internet of things. Wirel Pers Commun
of medical things with blockchain cybersecurity. J Supercomput
130(4):2959–2980. ​h​t​t​p​​s​:​/​​/​d​o​i​​.​o​​r​g​/​​1​0​.​1​​0​0​7​​/​s​1​​1​2​7​7​-​0​2​3​-​1​0​4​1​1​-​6
81(1):1–22. ​h​t​t​p​​s​:​/​​/​d​o​i​​.​o​​r​g​/​​1​0​.​1​​0​0​7​​/​s​1​​1​2​2​7​-​0​2​4​-​0​6​7​8​2​-​7
23. Saied M, Guirguis S, Madbouly M (2024) Review of artificial
7. Khan AA et al (2025) Artificial Intelligence, Internet of Things,
intelligence for enhancing intrusion detection in the internet of
and Blockchain Empowering Future Vehicular Developments:
things. Eng Appl Artif Intell 127:107231. ​h​t​t​p​​s​:​/​​/​d​o​i​​.​o​​r​g​/​​1​0​.​1​​0​1​6​​
A Comprehensive Multi-Hierarchical Lifecycle Review,
/​J​.​​E​N​G​​A​P​P​​A​I​.​2​​0​2​​3​.​1​0​7​2​3​1
Human-centric Comput. Inf. Sci., vol. 15, no. March, pp. 1–19.​
24. Issa MM, Aljanabi M, Muhialdeen HM (2024) Systematic litera-
h​t​t​p​​s​:​/​​/​d​o​i​​.​o​​r​g​/​​1​0​.​2​​2​9​6​​7​/​H​​C​I​S​.​2​0​2​5​.​1​5​.​0​1​3
ture review on intrusion detection systems: research trends, algo-
8. Khan AA (2025) A lightweight scalable hybrid authentication
rithms, methods, datasets, and limitations. J Intell Syst. ​h​t​t​p​​s​:​/​​/​d​o​
framework for internet of medical things (IoMT) using block-
i​​.​o​​r​g​/​​1​0​.​1​​5​1​5​​/​j​i​​s​y​s​-​2​0​2​3​-​0​2​4​8
chain hyperledger consortium network with edge computing. Sci
25. Mishra N, Mishra S (2024) A review of machine learning-based
Rep. ​h​t​t​p​​s​:​/​​/​d​o​i​​.​o​​r​g​/​​1​0​.​1​​0​3​8​​/​s​4​​1​5​9​8​-​0​2​5​-​0​5​1​3​0​-​w
intrusion detection system. EAI Endorsed Trans Internet Things
9. Feng L, Du J, Fu C, Song W (2023) Image Encryption Algorithm
10:1–8. ​h​t​t​p​​s​:​/​​/​d​o​i​​.​o​​r​g​/​​1​0​.​4​​1​0​8​​/​e​e​​t​i​o​t​.​5​3​3​2
Combining Chaotic Image Encryption and Convolutional Neural
26. Kumar D, Kumar M (2024) Hybrid cryptographic approach for data
Network, Electron. Vol. 12, Page 3455, vol. 12, no. 16, p. 3455,
security using elliptic curve cryptography for IoT. Int J Comput Netw
Aug. 2023. ​h​t​t​p​​s​:​/​​/​d​o​i​​.​o​​r​g​/​​1​0​.​3​​3​9​0​​/​E​L​​E​C​T​R​O​N​I​C​S​1​2​1​6​3​4​5​5
Inf Secur 16(2):42–54. ​h​t​t​p​​s​:​/​​/​d​o​i​​.​o​​r​g​/​​1​0​.​5​​8​1​5​​/​i​j​​c​n​i​s​.​2​0​2​4​.​0​2​.​0​4
10. Wang Z, Fok KW, Thing VLL (2022) Machine learning for
27. Bhandiwad V, Ragha LK (2025) Enhancing the security of IOT
encrypted malicious traffic detection: approaches, datasets and
enabled systems using light weight hybrid cryptography models.
comparative study. Comput Secur 113:102542. ​h​t​t​p​​s​:​/​​/​d​o​i​​.​o​​r​g​/​​1​
Cluster Comput. ​h​t​t​p​​s​:​/​​/​d​o​i​​.​o​​r​g​/​​1​0​.​1​​0​0​7​​/​s​1​​0​5​8​6​-​0​2​4​-​0​4​8​9​7​-​3
0​.​1​​0​1​6​​/​J​.​​C​O​S​E​.​2​0​2​1​.​1​0​2​5​4​2
28. Mousavi SK, Ghaffari A, Besharat S, Afshari H (2021) Security
11. Singh S, Sharma PK, Moon SY, Park JH (Feb. 2024) Advanced
of internet of things using RC4 and ECC algorithms (case study:
lightweight encryption algorithms for IoT devices: survey, challenges
smart irrigation systems). Wirel Pers Commun 116(3):1713–
and solutions. J Ambient Intell Humaniz Comput 15(2):1625–1642.​
1742. ​h​t​t​p​​s​:​/​​/​d​o​i​​.​o​​r​g​/​​1​0​.​1​​0​0​7​​/​s​1​​1​2​7​7​-​0​2​0​-​0​7​7​5​8​-​5
h​t​t​p​​s​:​/​​/​d​o​i​​.​o​​r​g​/​​1​0​.​1​​0​0​7​​/​S​1​​2​6​5​​2​-​0​​1​7​-​0​​4​9​​4​-​4​/​T​A​B​L​E​S​/​7
29. Ferdush J, Begum M, Uddin MS (2021) Chaotic lightweight
12. Jammula M, Vakamulla VM, Kondoju SK (Dec. 2022) Hybrid
cryptosystem for image encryption. Adv Multimedia. ​h​t​t​p​s​:​​​/​​/​d​o​​
lightweight cryptography with attribute-based encryption
i​.​o​​r​​g​​/​​1​0​​.​1​1​​​5​5​/​​2​​0​2​1​/​5​5​2​7​2​9​5
standard for secure and scalable IoT system. Conn Sci
34(1):2431–2447 ;WGROUP:STRING:PUBLICATION
13. Zhang L, Wang L (2024) A hybrid encryption approach for effi- Publisher’s Note Springer Nature remains neutral with regard to
cient and secure data transmission in IoT devices. J Eng Appl Sci jurisdictional claims in published maps and institutional affiliations.
71(1):1–18. ​h​t​t​p​​s​:​/​​/​d​o​i​​.​o​​r​g​/​​1​0​.​1​​1​8​6​​/​s​4​​4​1​4​7​-​0​2​4​-​0​0​4​5​9​-​x
14. Gharat NN, Jolly L (2024) Hybrid lightweight cryptography Springer Nature or its licensor (e.g. a society or other partner) holds
using AES and ECC for IoT security. Lect Notes Networks Syst exclusive rights to this article under a publishing agreement with the
896:241–258. ​h​t​t​p​​s​:​/​​/​d​o​i​​.​o​​r​g​/​​1​0​.​1​​0​0​7​​/​9​7​​8​-​9​8​1​-​9​9​-​9​8​1​1​-​1​_​1​9 author(s) or other rightsholder(s); author self-archiving of the accepted
15. Ahmed KI, Tahir M, Habaebi MH, Lau SL, Ahad A (2021) manuscript version of this article is solely governed by the terms of
Machine learning for authentication and authorization in IoT: such publishing agreement and applicable law.

13
 307 Page 18 of 18 Peer-to-Peer Networking and Applications (2025) 18:307

Geeta Sharma Geeta Sharma Sanjeev Jain Sanjeev Jain

received the master’s degree from (Senior Member, IEEE) was born
Kurukshetra University, Kuruk- in Vidisha, in 1967. He received
shetra. She is currently pursuing the [Link]. degree in electronics
the Ph.D. degree with the Depart- from the Samrat Ashok Techno-
ment of Computer Science and logical Institute, Vidisha, the
Information Technology, Central [Link]. degree in computer engi-
University of Jammu, Jammu, neering from IIT Delhi, in 1992,
India. Her research interests and the Ph.D. degree from Bar-
include learning-based naviga- katullah Vishwavidyalaya, India,
tion, path planning of UAVs, and in 2000. Currently, he is the
deep reinforcement learning. Visionary Vice-Chancellor at the
Central University of Jammu,
Jammu, India, where his leader-
ship is shaping the future of educa-
tion. His illustrious career path includes a tenure as the Director at
Dr. Narinder Verma Dr. Indian Institute of Information Technology, Design and Manufacturing,
Narinder Verma is currently Jabalpur, an institute of national importance, established by the Govern-
working as an Assistant Professor ment of India, from 2019 to 2021. Prior to that, he steered the helm as
in the School of Computer Science the Vice-Chancellor at Shri Mata Vaishno Devi University, Katra, from
and Engineering, IILM University, 2016 to 2019, leaving an indelible mark of academic excellence. His
Greater Noida, India. He has done leadership acumen is further exemplified by his tenure as the Director
his Ph.D. in Computer Science of the Madhav Institute of Technology and Science, Gwalior, Madhya
and IT at the Central University of Pradesh, from 2010 to 2016. His illustrious career spans over three
Jammu, India, with a research decades, including six impactful years as the Director of MITS, a
focus on cybersecurity, IoT, Grant-in-Aid Autonomous Engineering College, Government of Mad-
machine learning, and cyber- hya Pradesh, where he demonstrated unparalleled dedication to aca-
physical systems. He has over two demic advancement and institutional growth.
years of experience as a Senior
Research Fellow (SRF) on a ₹10- Radhe Shyam Sharma Radhe
crore DRDO-sponsored project Shyam Sharma (Member, IEEE)
and three years of teaching experience at the University of Jammu received the master’s degree in
(Kishtwar Campus). Dr. Verma has authored and co-authored over 20 modeling and simulation from the
publications, including journal articles, conference papers, and book Defence Institute of Advanced
chapters in SCI and Scopus-indexed Journals. He has also filed and Technology, Pune, India, in 2013,
published more than eight patents in the areas of IoT, anomaly and the Ph.D. degree in electrical
detection, and autonomous systems. He also served as a reviewer for engineering from Indian Institute
reputed conferences and journals. of Technology (IIT) Kanpur,
India, in 2020. He is currently an
Assistant Professor at IIT Mandi,
India. Before joining IIT Mandi,
he was an Autonomy Senior Engi-
neer at the Integrated Components
and Solutions Division, Caterpil-
lar. His research interests include robotics, vision, control, and auto-
mation, with an emphasis on visual servoing, robot learning from
demonstration, multi-robot systems, robot programming, and model-
ing and simulation.

13