0% found this document useful (0 votes)
14 views13 pages

Demerits of Keystroke Recognition Systems

The document discusses the limitations of unimodal biometric systems, particularly keystroke recognition, and introduces multimodal biometric systems that utilize multiple biometric inputs for improved reliability and security. It outlines the benefits of multimodal systems, including enhanced accuracy, liveliness detection, and the ability to handle failures of individual identifiers. Additionally, it details various integration strategies and performance evaluation criteria for biometric systems, emphasizing the importance of balancing false acceptance and rejection rates.
Copyright
© All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PDF, TXT or read online on Scribd
0% found this document useful (0 votes)
14 views13 pages

Demerits of Keystroke Recognition Systems

The document discusses the limitations of unimodal biometric systems, particularly keystroke recognition, and introduces multimodal biometric systems that utilize multiple biometric inputs for improved reliability and security. It outlines the benefits of multimodal systems, including enhanced accuracy, liveliness detection, and the ability to handle failures of individual identifiers. Additionally, it details various integration strategies and performance evaluation criteria for biometric systems, emphasizing the importance of balancing false acceptance and rejection rates.
Copyright
© All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PDF, TXT or read online on Scribd

BIOMETRICS SYSTEM

Demerits of Keystroke Recognition System

 The candidate’s typing rhythm can change between a number of days or within a day
itself because of tiredness, sickness, influence of medicines or alcohol, change of
keyboard, etc.
 There are no known features dedicated solely to carry out discriminating information.

TOPIC 3: INTRODUCTION TO MULTIMODAL BIOMETRIC SYSTEM

DESCRIPTION:

All the biometric systems we discussed till now were unimodal, which take single source of
information for authentication. As the name depicts, multimodal biometric systems work on
accepting information from two or more biometric inputs.

A multimodal biometric system increases the scope and variety of input information the system
takes from the users for authentication.

Why Multimodal Biometrics is required?

The unimodal systems have to deal with various challenges such as lack of secrecy, non-
universality of samples, extent of user’s comfort and freedom while dealing with the system,
spoofing attacks on stored data, etc.

Some of these challenges can be addressed by employing a multimodal biometric system.

There are several more reasons for its requirement, such as −

 Availability of multiple traits makes the multimodal system more reliable.


 A multimodal biometric system increases security and secrecy of user data.
 A multimodal biometric system conducts fusion strategies to combine decisions from
each subsystem and then comes up with a conclusion. This makes a multimodal system
more accurate.
BIOMETRICS SYSTEM

 If any of the identifiers fail to work for known or unknown reasons, the system still can
provide security by employing the other identifier.
 Multimodal systems can provide knowledge about “liveliness” of the sample being
entered by applying liveliness detection techniques. This makes them capable to detect
and handle spoofing.

Working of Multimodal Biometric System

Multimodal biometric system has all the conventional modules a unimodal system has −

 Capturing module
 Feature extraction module
 Comparison module
 Decision making module

In addition, it has a fusion technique to integrate the information from two different
authentication systems. The fusion can be done at any of the following levels −

 During feature extraction.


 During comparison of live samples with stored biometric templates.
 During decision making.

Page 10
BIOMETRICS SYSTEM

The multimodal biometric systems that integrate or fuse the information at initial stage are
considered to be more effective than the systems those integrate the information at the later
stages. The obvious reason to this is, the early stage contains more accurate information than the
matching scores of the comparison modules.

Fusion Scenarios in Multimodal Biometric System

Within a multimodal biometric system, there can be variety in number of traits and components.
They can be as follows −

 Single biometric trait, multiple sensors.


 Single biometric trait, multiple classifiers (say, minutiae-based matcher and texture-based
matcher).
 Single biometric trait, multiple units (say, multiple fingers).
 Multiple biometric traits of an individual (say iris, fingerprint, etc.).

These traits are then operated upon to confirm user’s identity.

Design Issues with Multimodal Biometric Systems

You need to consider a number of factors while designing a multimodal biometric system

 Level of security you need to bring in.


 The number of users who will use the system.
 Types of biometric traits you need to acquire.
 The number of biometric traits from the users.
 The level at which multiple biometric traits need integration.
 The technique to be adopted to integrate the information.
 The trade-off between development cost versus system performance.

Page 11
TOPIC 4: Integration strategies

Description:

NEED FOR BIOMETRICS INTEGRATION

 Unimodal biometric systems suffer many drawbacks problem of noisy data, intra class
variation, improper, User sensor adjustment, interclass similarities insufficient population
coverage and spoof attacks.
 These problems lead to higher false rejection Rate (FRR) and false Acceptance Rate
FAR).
 Unimodal biometric system may not be able to achieve 100% performance. Multi
biometric systems are comparatively found to be more reliable due to integration of
multiple, independent biometrics information.
 Moreover, they help in resolving various problems encountered by Unimodal biometric
systems. They indexing or filtering of large biometric databases and are robust to noise,
provide universal coverage and improve matching accuracy.
 Further, when two or more modalities are used for authentication, it becomes difficult to
spoof the biometric system.
 Experiments have shown that the accuracy of multimodality can reach 100% and its
performance is far better than Unimodal identification
INTEGRATIONLEVEL’S
Information integration in a multi biometric system can be done at four different levels

A. Integration at feature extraction level:


 Feature sets extracted from biometric traits are integrated and the combined feature
vector thus obtained is passed to the matching module.

B Image level integration


 Each image representation in bit form is integrated to obtain a single bit sequence
representing the final image

Page 12
[Link] score level
 Matching score values, representing the similarities between biometric information
obtained and the already stored biometric templates for each modality are combined.
 Applying fusion at this level is preferred as it is easy to obtain and combine matching
scores.

D. Decision level
 Separate authentication process is carried out for each biometric modality and the
decisions taken by individual systems are integrated to obtain the final result.
 Integration at this level is supposed to be rigid because of availability of limited
information.

 Integrating information at an early stage is believed to be more effective since the feature
set contains more information about the input data.
 An effective fusion scheme is the key to a successful multi biometric system. Fusion
rules must be chosen according to the type of application, biometric traits to be used and
the level of fusion.
 Different matching algorithms and several rules are then applied to the Information
obtained for reaching at a decision.
TOPIC 5: Architecture of Multimodal biometrics

Page 13
Page 14
TOPIC 6: LEVEL OF FUSION

Page 15
Fusion in multimodal biometric systems

Page 16
 We use more than one biometric modality in multimodal biometric systems and hence we
have more than one decision channels.
 Thus arises the need to design a mechanism which can combine the classification
outcome from each biometric channel and this mechanism is known as biometric fusion.
 This fusion combines the measurements from different biometric attributes to enhance
the strengths and decrease the weaknesses of the individual measurements.
 Fusion can be used to address a number of issues faced in implementation of biometric
systems such as accuracy, efficiency, robustness, applicability and universality.
 There are various levels of fusing the biometric traits which can be used to increase
robustness of the multimodal biometric system.
 They are sensor level fusion, feature level fusion, matching score level fusion and
decision level fusion.

Sensor level fusion

 In sensor level fusion, we fuse the biometric traits coming from different sensors such as
fingerprint scanner, iris scanner, video camera etc. to form a merged biometric trait and
process.

Feature level fusion

Page 17
 In feature level fusion, signals coming from different biometric channels are first
processed after which the feature vectors are extracted separately from each biometric
trait.
 The feature vectors are then combined to form a composite feature vector using a specific
fusion algorithm and then used for further classification.
 In feature level fusion, some reduction techniques need to be used in order to select only
the useful features.
 Features contain richer information of biometric traits as compared to matching score or
decision of matcher and thus fusion at the feature level provides better recognition
results.
 It has also been observed that feature level fusion provides more accuracy when the
features of different biometric modalities are compatible with each other.

Matching score level fusion

 In this fusion level, the feature vectors are processed separately rather than combining
them. Then an individual matching score is found and based on the accuracy of each
biometric channel, we then fuse the matching level to find a composite matching score
which will be used for classification.
 We can use various techniques such as logistic regression, highest rank, Bayes rule, mean
fusion etc. to combine match scores.
 In addition to this, another important aspect of this fusion is the normalization of scores
acquired from different modalities.
 We can use techniques such as Min-max, z-score, piecewise linear etc. to achieve
normalization of the match scores. Matching score level fusion has lesser complexity
than the other fusion levels and hence it is widely used.

Decision level fusion

 In decision level fusion, each biometric trait is first pre-classified separately.


 The individual biometric trait is first captured and then features are extracted from the
captured trait. The traits are classified as either accept or reject based on these extracted
features.

Page 18
 The final classification is obtained by combining the outputs of different modalities.

Topic 7: Performance evaluation

Description

Criteria for Effective Biometric System

There are seven basic criteria for measuring effectiveness of a biometric system −

 Uniqueness − It determines how uniquely a biometric system can recognize a user from a
group of users. It is a primary criterion.
 Universality − It indicates requirement for unique characteristics of each person in the
world, which cannot be reproduced. It is a secondary criterion.
 Permanence − It indicates that a personal trait recorded needs to be constant in the
database for a certain time period.
 Collectability − It is the ease at which a person’s trait can be acquired, measured, or
processed further.
 Performance − It is the efficiency of system in terms of accuracy, speed, fault handling,
and robustness.
 Acceptability − It is the user-friendliness, or how good the users accept the technology
such that they are cooperative to let their biometric trait captured and assessed.
 Circumvention − It is the ease with which a trait is possibly imitated using an artifact or
substitute.

Biometric System Performance

Biometric system manufacturers claim high system performance which is practically difficult to
achieve in actual operating environments. The possible reasons are, tests conducted in controlled
environment setups, limitations on hardware, etc.

For example, a voice recognition system can work efficiently only in quiet environment, a facial
recognition system can work fine if lighting conditions are controlled, and candidates can be
trained to clean and place their fingers properly on the fingerprint scanners.

Page 19
However, in practice, such ideal conditions may not be available in the target operating
environment.

Performance Measurements

The performance measurements of a biometric system are closely tied to False Reject Rate
(FRR) and False Accept Rate (FAR).

FRR is also known as Type-I error or False Non Match Rate (FNMR) which states the
likelihood of a legitimate user being rejected by the system.

FAR is referred to as Type-II error or False Match Rate (FMR) which states the likelihood of a
false identity claim being accepted by the system.

An ideal biometric system is expected to produce zero value for both FAR and FRR. Means it
should accept all genuine users and reject all fake identity claims, which is practically not
achievable.

FAR and FRR are inversely proportional to each other. If FAR is improved, then the FRR
declines. A biometric system providing high FRR ensures high security. If the FRR is too high,
then the system requires to enter the live sample a number of times, which makes it less efficient.

The performance of current biometrics technologies is far from the ideal. Hence the system
developers need to keep a good balance between these two factors depending on the security
requirements.

False Acceptance Rate (FAR)


 The FAR is the frequency that a non authorized person is accepted as authorized.
 Because a false acceptance can often lead to damages, FAR is generally a security
relevant measure.
 FAR is a non-stationary statistical quantity which does not only show a personal
correlation, it can even be determined for each individual biometric characteristic (called
personal FAR).
False Rejection Rate (FRR)

Page 20
 The FRR is the frequency that an authorized person is rejected access.
 FRR is generally thought of as a comfort criteria, because a false rejection is most of all
annoying.
 FRR is a non-stationary statistical quantity which does not only show a strong personal
correlation, it can even be determined for each individual biometric characteristic (called
personal FRR).
Failure To Enrol rate (FTE, also FER)
 The FER is the proportion of people who fail to be enrolled successfully.
 FER is a non-stationary statistical quantity which does not only show a strong personal
correlation, it can even be determined for each individual biometric characteristic (called
personal FER).
Those who are enrolled yet but are mistakenly rejected after many
verification/identification attempts count for the Failure To Acquire (FTA) rate.
 FTA can originate through temporarily not measurable features ("bandage", non-
sufficient sensor image quality, etc.).
 The FTA usually is considered within the FRR and need not be calculated separately, see
also FNMR and FMR.
False Identification Rate (FIR)
 The False Identification Rate is the probability in an identification that the biometric
features are falsely assigned to a reference.
 The exact definition depends on the assignment strategy; namely, after feature
comparison, often more than one reference will exceed the decision threshold.
Equal Error Rate, or EER
 It is the measure of the likelihood that the biometric security system has the FAR equal
to the FRR. This is a very important measure for any biometrics system.
 The reason why this can occur will be discussed in later articles.

Topic 8: Memory Requirement and allocation

Description:

Page 21

Common questions

Powered by AI

Multimodal biometric systems address several limitations of unimodal systems by increasing reliability and security through the use of multiple biometric traits. This reduces the impact of issues like noise, intra-class variation, and spoofing attacks . The integration of different modalities helps resolve problems such as noisy data and insufficient population coverage, and improves matching accuracy by combining independent biometric information . Furthermore, multimodal systems enhance performance by leveraging fusion strategies that combine decisions from multiple channels, thereby making them more robust to variations and attacks compared to unimodal systems .

Integration strategies in biometrics include feature extraction-level integration, matching score-level integration, and decision-level integration . Choosing the right strategy is crucial because each level has different implications for data richness, computational complexity, and accuracy. Feature level integration, for instance, maintains comprehensive data but requires advanced processing capabilities. Matching score-level integration is simpler to implement, optimizing both accuracy and efficiency with appropriate normalization techniques. Decision-level integration offers flexibility by combining outputs of multiple modalities, though at the expense of finer data granularity. Successful biometric systems tailor integration strategies to application-specific demands for optimal performance .

Multimodal biometric systems are considered more secure against spoofing than unimodal systems because they utilize multiple independent biometric traits for verification . This diversification makes it more difficult for attackers to spoof as all biometric trait inputs must simultaneously be falsified to breach the system. Additionally, these systems have the capability to analyze multiple data sources and use liveliness detection techniques, further mitigating spoofing potential . The combination of various modalities necessitates that any spoofing attack is adept at mimicking multiple authentic features and not just a single source, thereby enhancing security significantly .

Fusion levels in multimodal biometric systems can be organized as sensor level, feature level, matching score level, and decision level fusions. Sensor level fusion combines data from different sensors to form a unified trait for processing, which can increase initial robustness . Feature level fusion integrates feature vectors to form a comprehensive vector for classification, yielding higher accuracy due to the richness of data at this stage . Matching score level fusion processes scores separately and then combines them, offering computational simplicity and ease of implementation . Decision level fusion combines outputs from separate pre-classified traits, providing a mechanism that can potentially accommodate variations in individual modalities .

Early stage fusion in multimodal biometric systems is considered more effective because this stage involves more accurate and rich data. By integrating information at the feature extraction level, the system utilizes detailed and multiple sources of input data before any transformation or score-based comparisons, which reduces information loss and enhances the precision of the final decision . Furthermore, early fusion helps create a composite feature vector that can be processed more robustly, often leading to better recognition results .

Current biometric technologies face challenges in achieving ideal performance levels due to environmental factors and the variability of operating conditions. Systems are often tested in controlled environments where factors such as lighting and noise can be managed, but real-world applications encounter uncontrolled conditions, leading to decreased performance . Additionally, the inverse relationship between False Acceptance Rate (FAR) and False Rejection Rate (FRR) makes it difficult to simultaneously minimize both, which is crucial for achieving ideal performance. The non-stationary nature of biometric system metrics, like FAR and FRR, also complicates maintaining consistently high performance across different dynamics and user populations .

The False Acceptance Rate (FAR) and False Rejection Rate (FRR) are inversely proportional in biometric systems, meaning an improvement in one often results in the degradation of the other . This relationship significantly influences design as it requires a balanced approach to optimize system security (low FAR) while maintaining user convenience (low FRR). High security demands low FAR, but if FRR becomes too high, legitimate users experience frequent rejections, decreasing efficiency. Thus, developers must carefully calibrate these rates based on specific security needs and user scenarios .

Biometric systems are evaluated based on criteria such as uniqueness, universality, permanence, collectability, performance, acceptability, and circumvention . These criteria impact the system's effectiveness by ensuring comprehensive coverage (uniqueness and universality), consistent recognition over time (permanence), user friendliness (collectability and acceptability), and resistance to spoofing (circumvention). Performance directly influences system accuracy, speed, and fault handling abilities, thus determining overall reliability and efficiency .

Matching score level fusion benefits the implementation of multimodal biometric systems by offering a relatively simple and effective way to increase system accuracy . It allows the system to process biometric traits separately, finding individual matching scores for each and then combining these using techniques like logistic regression or Bayes rule to obtain a comprehensive score. This method reduces computational complexity compared to feature level fusion and supports various normalization techniques to handle discrepancies in scores from different modalities . Additionally, it helps improve robustness by allowing easy integration of new biometric modalities without extensive restructuring .

When designing a multimodal biometric system, key considerations include the level of security required, the number of users, types and number of biometric traits to acquire, and the level and technique of integration of these traits . Additionally, design must balance development costs against system performance, ensuring that the system is efficient and effective while remaining within budget constraints .

You might also like