1

PROJECT - I REPORT
ON
Enhancing Biometric Template Security through a
Cancelable System
SUBMITTED IN PARTIAL FULFILMENT OF THE REQUIREMENTS FOR THE
AWARD OF DEGREE OF

BACHELOR OF TECHNOLOGY
IN
COMPUTER SCIENCE & ENGINEERING

Submitted By:

Name: Vikas Kashyap Ankit Kumar Akash Kumar

Roll No. 22becse46 22becse06 22becse04

Group No. 22
Under the supervision of
Dr. Deepti Malhotr
(Assistant Professor)

DEPARTMENT OF COMPUTER SCIENCE & ENGINEERING

CENTRAL UNIVERSITY OF JAMMU
Rahya - Suchani (Bagla), District Samba-181143, Jammu (J&K)
 2

Abstract
In today’s rapidly advancing digital landscape, ensuring the security and privacy of personal identity
has become one of the most pressing concerns. Conventional authentication mechanisms, including
passwords, tokens, and PINs, are increasingly vulnerable to theft, duplication, and misuse. With the
growth of cyber threats and identity fraud, such approaches often fail to guarantee long-term reliability.
Biometric authentication has therefore emerged as a more effective and convenient solution, as it relies
on unique physiological or behavioral traits such as fingerprints, iris patterns, palm veins, or facial
features. These traits are difficult to forge or lose, making biometrics inherently more secure than
knowledge-based or token-based systems. However, the irreversible nature of biometric data introduces
new challenges. Once a biometric template is compromised, it cannot be replaced or reissued, unlike
passwords or smart cards. Centralized storage of raw biometric templates further exposes sensitive
personal information to large-scale breaches, identity theft, and privacy violations.
To overcome these challenges, this project focuses on the development of a Cancelable Biometric
System that secures biometric data through the application of intentional, irreversible transformations.
Cancelable biometrics aim to provide three key properties: non-invertibility, where transformed data
cannot be used to reconstruct the original template; revocability, where compromised templates can be
reissued without requiring the user to change their inherent biometric traits; and diversity, where
different transformed templates can be generated for different applications, thereby preventing cross-
matching and misuse of personal data. By incorporating these principles, cancelable biometrics
strengthen trust in authentication systems and mitigate the risks associated with traditional approaches.
The scope of this project involves an in-depth study of existing template protection methods such as
Random Projection, BioHashing, Bloom Filters, and advanced approaches leveraging deep learning and
hybrid cryptographic schemes. Based on this analysis, a secure and robust cancelable biometric
framework will be designed and implemented. The framework will be tested using publicly available
biometric datasets, focusing on modalities such as fingerprints, iris, or palmprints. Its performance will
be evaluated against standard metrics like False Acceptance Rate (FAR), False Rejection Rate (FRR),
and Equal Error Rate (EER). Special emphasis will be placed on measuring resistance against template
inversion and similarity-based attacks, which represent critical vulnerabilities in current systems.
The proposed system offers several advantages. It preserves accuracy comparable to conventional
biometric systems while providing enhanced privacy and protection against unauthorized access. It
enables the revocation and reissuance of templates in the event of compromise and ensures scalability
by supporting multiple biometric modalities. Moreover, its design emphasizes computational efficiency,
making it suitable for integration into real-world applications such as secure access control, digital
identity management, e-governance, banking, and healthcare systems.
By addressing the shortcomings of existing biometric systems, this research contributes significantly to
the field of biometric template protection. It provides a practical framework that balances security,
privacy, revocability, and usability, thereby making biometric authentication systems more secure,
adaptable, and future-ready. The successful implementation of this project will not only enhance user
 3

trust in biometric technologies but also serve as a foundation for further advancements in privacy-
preserving authentication in an increasingly digital world.
 4

Acknowledgement
We would like to express our sincere gratitude to all those who have contributed directly or indirectly to
the successful completion of this project report titled “Enhancing Biometric Template Security through a
Cancelable System.”
First and foremost, we are deeply grateful to Dr. Deepti Malhotra, Assistant Professor, Department of
Computer Science & Engineering, Central University of Jammu, for her valuable guidance,
encouragement, and continuous support throughout the course of this work. Her insightful suggestions,
timely feedback, and constant motivation have been instrumental in shaping this project and ensuring its
successful completion.
We also wish to extend our thanks to the faculty members of the Department of Computer Science &
Engineering, Central University of Jammu, for their cooperation and for providing us with the knowledge
and resources necessary to carry out this project.
Our heartfelt appreciation goes to our project team members, whose collaboration, dedication, and
consistent efforts made it possible to successfully implement and document this work. The teamwork and
mutual support have been invaluable in overcoming challenges during the project.
We would also like to acknowledge our friends and classmates for their encouragement and constructive
feedback, which helped us refine our ideas and improve the overall quality of the work.
Last but not least, we remain indebted to our parents and families for their constant encouragement,
patience, and moral support, without which this achievement would not have been possible.
We gratefully acknowledge all contributions that have made this project a reality.
 5

TABLE OF CONTENTS

PAGE NO.
CANDIDATE'S DECLARATION I
TABLE OF CONTENTS II
LIST OF FIGURES III
LIST OF TABLES IV
1. ABSTRACT 1
2. INTRODUCTION 2-5
3. LITERATURE SURVEY 6-10
4. CHALLENGES AND RESEARCH GAPS 11
5. PROBLEM FORMULATION 12-13
6. RESEARCH OBJECTIVE 14
7. METHODOLOGY/ PLANNING OF WORK 15
8. CONCLUSION 16
9. REFERENCES 17-19
 3

LIST OF FIGURES

FIGURE NO. NAME OF FIGURE PAGE NO.

FIG 1 SECURITY SYSTEM 2
FIG 2 TYPES OF BIOMETRICS 3
FIG 3 CHALLENGES WITH SECURITY SYSTEMS 3
FIG 4 TAXONOMY FOR CANCELABLE BIOMETRIC 4
TEMPLATE GENERATION
FIG 5 PRISMA DIAGRAM 5
FIG 6 PROPOSED METHODOLOGY 15
 4

LIST OF TABLES

TABLE NO. NAME OF TABLE PAGE NO.

TABLE 1 STATISTICAL ANALYSIS OF CANCELABLE BIOMETRIC TEMPLATE
7
GENERATION USING RANDOM PROJECTION
TABLE 2 STATISTICAL ANALYSIS OF CANCELABLE BIOMETRIC TEMPLATE
8
GENERATION USING BLOOM FILTER
TABLE 3 STATISTICAL ANALYSIS OF CANCELABLE BIOMETRIC TEMPLATE
9
GENERATION USING BIOHASHING
TABLE 4 STATISTICAL ANALYSIS OF CANCELABLE BIOMETRIC TEMPLATE
10
GENERATION USING DEEP LEARNING
 3

1.1 Project Overview

The project titled “Enhancing Biometric Template Security through a Cancelable System” falls
under the category of Research-Based Application Development. It primarily focuses on strengthening
biometric authentication systems by integrating cancelable biometric techniques to safeguard sensitive
identity data.

In the current digital landscape, where cyber threats and data breaches are on the rise, biometric systems
such as fingerprints, iris scans, and facial recognition have become integral for secure authentication.
Unlike passwords or PINs, biometric traits cannot be forgotten or easily duplicated, making them reliable
for identity verification. However, conventional biometric systems face critical challenges: once a
biometric template is compromised, it cannot be revoked or replaced like a password. This leads to
severe security and privacy risks.
To address these issues, the proposed project emphasizes the concept of Cancelable Biometrics—a
method that applies irreversible transformations to biometric templates. By generating protected,
revocable, and application-specific templates, cancelable systems provide multiple layers of security.
Even if a transformed template is leaked, the original biometric data remains unrecoverable. This ensures
revocability, diversity, and non-invertibility, which are essential to building user trust and long-term
reliability in biometric authentication.

The research will involve analyzing existing transformation techniques such as BioHashing, Random
Projection, and Bloom Filter approaches, followed by the design and development of a secure
cancelable biometric framework. The system will be evaluated on the basis of accuracy, robustness
against inversion attacks, and revocability.

This project not only advances theoretical understanding in the field of biometric cryptosystems but also
proposes a practical system design that can be integrated into real-world applications like smartphones,
e-banking, and access control systems. Ultimately, the project aims to contribute towards developing
privacy-preserving, scalable, and secure authentication systems, aligning with the increasing global
demand for data security and digital trust.
 4

1.2 Objectives of the Project

The main objectives of this project are:

1. To analyze and study different cancelable biometric template generation techniques for
enhancing the security and privacy of biometric data.

2. To design and develop a secure cancelable biometric framework that ensures template
revocability and non-invertibility.

3. To implement and evaluate the proposed system using benchmark biometric datasets,
measuring performance in terms of accuracy, FAR (False Acceptance Rate), FRR (False
Rejection Rate), and resistance to attacks.

4. To compare and validate the proposed framework against existing methodologies, identifying
improvements in robustness, scalability, and privacy preservation.

1.3 Problem Formulation

Biometric authentication has become a cornerstone of digital security, yet traditional systems are
vulnerable. Once compromised, a biometric template cannot be changed like a password, exposing users
to lifelong identity risks. Storing raw templates in centralized databases further increases the chances of
breaches. Hence, the challenge is to develop a cancelable biometric system that generates secure,
revocable, and non-invertible templates without degrading recognition accuracy. The problem is to
balance security, privacy, and system performance within a practical and scalable framework.

1.4 Existing System

Existing biometric systems store raw or partially protected templates, making them highly vulnerable to
theft and misuse. Though methods like encryption and biometric cryptosystems exist, they often
suffer from performance trade-offs, high computational overhead, or lack of revocability. Some
cancelable biometric approaches (e.g., Random Projection, BioHashing, and Bloom Filters) have been
proposed, but they either compromise accuracy or fail to withstand advanced inversion and similarity-
based attacks. Thus, existing systems are not fully secure, scalable, or adaptable to diverse real-world
applications.
 5

1.5 Proposed System

The proposed system introduces a Cancelable Biometric Framework where original biometric traits
(fingerprint, iris, or face) undergo irreversible transformations to generate secure templates. These
templates ensure:

 Revocability – A new template can be issued if compromised.

 Diversity – Different applications can use different transformed templates.

 Non-invertibility – Original biometric traits cannot be reconstructed from stolen data.

The system leverages advanced transformation techniques (e.g., improved Random Projection,
BioHashing with deep learning, and Bloom Filter methods) and evaluates them against accuracy,
robustness, and resistance to attacks.

1.6 Features of the Project

 High Security: Protects biometric templates using irreversible transformations.

 Revocability: Enables re-issuance of new templates in case of compromise.

 Scalability: Adaptable to different biometric modalities (fingerprint, iris, face).

 Accuracy Maintenance: Ensures minimal performance loss while protecting privacy.

 Robustness: Provides resistance to inversion, similarity-preserving, and spoofing attacks.

 Application Versatility: Suitable for use in smartphones, banking, e-governance, and access
control.
 6

Flow chart
 7

Use case
 8

2.1 Feasibility Study

A feasibility study evaluates whether the proposed cancelable biometric system can be developed
effectively in terms of technical, operational, and economic aspects:

 Technical Feasibility:
The project is technically feasible as it uses existing biometric datasets (fingerprint/iris/face) and
applies transformation techniques such as BioHashing, Random Projection, and Bloom Filter.
Implementation can be carried out using Python, MATLAB, or similar platforms with machine
learning and image-processing libraries. Required computational resources (high-performance
PCs/servers) are readily available.

 Operational Feasibility:
The system enhances the security of biometric authentication, directly addressing real-world
concerns such as template theft and privacy risks. Since cancelable templates can be revoked and
re-issued, the solution is practical for applications like mobile authentication, online banking, and
access control.

 Economic Feasibility:
The project requires minimal cost, as it relies on publicly available biometric datasets and open-
source software frameworks. Hardware requirements are modest, and no specialized biometric
devices beyond standard scanners are necessary. Hence, the project is economically viable for
research and deployment.

Thus, the proposed solution is feasible in all aspects and can be successfully implemented within the
academic and industry scope.
 9

2.2 Software Requirement Specification (SRS) Document

a. Functional Requirements

1. The system must accept biometric input (e.g., fingerprint/iris/face image).

2. It must generate a transformed cancelable biometric template using selected algorithms.

3. It must allow revocability (generate new template upon request).

4. The system should compare templates for authentication and return results (accept/reject).
5. The system must provide evaluation metrics such as accuracy, FAR, and FRR.

b. Non-Functional Requirements

1. Security – Irreversible transformation ensures non-invertibility.

2. Performance – Template generation and matching must execute in reasonable time (<2 seconds
for single input).

3. Usability – Simple interface for dataset input and result visualization.

4. Portability – The system should run on standard OS (Windows/Linux).

5. Scalability – Support multiple biometric modalities and datasets.

c. Software and Hardware Requirements

 Software: Python/Java/MATLAB, OpenCV, NumPy, TensorFlow/PyTorch, MySQL/SQLite for
storage.

 Hardware: Standard PC with minimum 8GB RAM, i5 processor, 500GB storage.

 10

2.3 SDLC Model Used

For this project, the Iterative Waterfall Model is selected. This model is suitable because:

1. The project involves research and development, where requirements and methods may evolve
during implementation.

2. Each phase (Requirement Analysis → Design → Implementation → Testing → Deployment) is

clearly defined, but feedback is possible between stages.

3. Iterations allow performance tuning of cancelable biometric techniques after initial testing.

4. It balances structured progress (like Waterfall) with adaptability (like Agile).

Thus, the Iterative Waterfall Model ensures both systematic development and flexibility for research-
driven improvements.
 11

3.1 Product Perspective

The proposed system, “Cancelable Biometric Template Protection Framework,” is designed as a
middleware security enhancement that fits into the broader ecosystem of biometric authentication
systems. In traditional systems, biometric data such as fingerprints, iris patterns, or facial images are
captured and stored in databases for verification. However, this storage of raw biometric data is highly
vulnerable to cyberattacks. If compromised, the biometric traits cannot be revoked or replaced, unlike
a password.

This project addresses that issue by embedding Cancelable Biometric Techniques into the
authentication workflow. The cancelable system sits between biometric feature extraction and
template storage/matching, ensuring that the original traits are never stored in their raw form. Instead,
irreversible transformations generate unique, revocable templates. These templates can vary from one
application to another, providing diversity and protecting against cross-matching across different
systems.

From a software development perspective, the system is modular in design. It consists of the
following components:

1. Input Module – accepts biometric data from users or datasets.

2. Preprocessing Unit – enhances and normalizes biometric data.

3. Transformation Engine – applies cancelable algorithms such as BioHashing, Random Projection,

or Bloom Filter.

4. Storage Module – stores transformed templates in a secure database.

5. Authentication Engine – compares new input templates with stored ones.

6. Revocation & Re-Issuance Module – generates new templates in case of compromise.

This layered perspective ensures separation of concerns, making the system both flexible and
extendable. In future, new transformation algorithms or storage protocols (e.g., blockchain-based
biometric storage) can be easily integrated.

Thus, the product perspective emphasizes that the system is not a replacement but an improvement of
existing biometric authentication systems, ensuring both security and usability.
 12

3.2 Product Functions

The cancelable biometric framework performs several essential functions that collectively strengthen
security, privacy, and usability:

1. Biometric Acquisition – The system captures input either from datasets or biometric sensors
(fingerprint scanner, iris scanner, or facial camera). This ensures compatibility with multiple
biometric modalities.
2. Data Preprocessing – Input images often contain noise due to environmental conditions (lighting,
angle, or sensor quality). Preprocessing steps such as noise filtering, normalization, and feature
extraction ensure reliable data for template generation.

3. Cancelable Transformation – At the heart of the system is the transformation process. Instead of
storing raw features, transformation algorithms (e.g., Random Projection, BioHashing, Bloom
Filter) are applied to generate protected templates. These are non-invertible and revocable,
ensuring that even if compromised, the original biometric cannot be reconstructed.

4. Template Storage – Transformed templates are securely stored in a database. Importantly, the
database contains no raw biometric data, minimizing risk in case of a breach.

5. Authentication/Verification – During login or verification, a new biometric sample is transformed

using the same algorithm and compared against stored templates. A match results in successful
authentication, otherwise access is denied.
6. Template Revocation & Reissuance – If a transformed template is compromised, the system can
generate a completely new template from the same biometric, much like resetting a password.

7. Performance Evaluation – The system continuously monitors metrics like FAR (False Acceptance
Rate), FRR (False Rejection Rate), and EER (Equal Error Rate) to evaluate security and usability.

These functions together create a robust authentication framework that not only prevents misuse of
biometric data but also maintains user convenience and system scalability.
 13

3.3 User Characteristics

The system is designed for multiple categories of users, each with different expectations and technical
capabilities:

 General End Users:

These are everyday individuals using biometric authentication for mobile apps, banking, e-
commerce, or workplace access. They prioritize ease of use, speed, and security. End users are
not expected to have technical expertise, so the interface must be intuitive, requiring minimal
interaction (just a scan).

 System Administrators:
Administrators are responsible for managing template databases, ensuring security compliance,
and updating algorithms when required. They should have technical knowledge of biometrics,
databases, and cybersecurity. Administrators also monitor system performance and respond to
breaches by revoking and reissuing templates.
 Researchers and Developers:
Since the project has a strong research orientation, it is also meant for academics and developers
experimenting with new cancelable biometric algorithms. They require flexible tools to test,
analyze, and compare performance metrics across different techniques.

 Stakeholders (Industry/Government):
Potential stakeholders include financial institutions, e-governance agencies, and smartphone
manufacturers who may adopt the technology. Their focus is on security, compliance, scalability,
and integration with existing infrastructure.
By identifying these characteristics, the system is tailored to meet both the non-technical needs of end
users and the technical requirements of administrators and researchers.
 14

3.4 Constraints
The system operates under several limitations that must be considered during design and
implementation:

1. Technical Constraints

o Limited computational power may affect real-time authentication performance when

using deep learning-based transformations.

o Image quality (low resolution, noisy captures) can reduce accuracy.

2. Security Constraints

o Transformation techniques must ensure non-invertibility; otherwise, attackers could

reconstruct original templates.

o Database encryption must be applied to secure stored templates.

3. Operational Constraints

o The system must perform authentication within 2 seconds per attempt to ensure usability.
o Revocability must not reduce matching accuracy significantly.

4. Legal and Ethical Constraints

o Biometric data must comply with privacy regulations like GDPR or India’s Personal Data
Protection Act.

o Unauthorized surveillance or misuse of biometric data must be prevented.

These constraints act as design boundaries, ensuring that the system remains both practical and
compliant while delivering high-level security.
 15

3.5 Use Case Model / Flow Chart / DFDs

The system’s workflow can be understood using Use Case Models and Data Flow Diagrams (DFDs).

Use Case Example: User Authentication

 Actors: User, System

 Steps:

1. User provides biometric input.

2. System preprocesses input and extracts features.

3. Cancelable transformation is applied.

4. Transformed template is compared with stored templates.

5. Authentication result (success/failure) is displayed.

Data Flow (Level 0 DFD)

 User → Input → Feature Extraction → Transformation Engine → Database (store template) /

Matching Engine (verify template) → Result.

These diagrams provide a visual understanding of how biometric input flows through the system.
 16

3.6 Database Design

The system database is designed to store only transformed templates instead of raw biometric data.
This provides security and ensures that even if the database is breached, no sensitive biometric
information is exposed.
 Entities: Users, Biometric Data, Transformed Templates.

 Relationships: Each user may have multiple biometric entries, each linked to one transformed
template.

 Storage Strategy:

o Templates stored as encrypted hash values.

o Metadata fields (algorithm used, timestamp) for auditing.

A normalized database design minimizes redundancy and ensures efficient retrieval during
authentication.
 17

3.7 Table Structure

Table: User_Auth

Field Name Data Type Description

User_ID INT (PK) Unique identifier for user

Name VARCHAR(50) User’s full name

Email VARCHAR(100) User’s email for contact

Biometric_Type VARCHAR(20) Finger vein

Cancelable_Temp TEXT Irreversible transformed template

Algorithm_Used VARCHAR(50) Transformation method applied

Timestamp DATETIME Date and time of enrollment/authentication

 18

3.8 ER Diagram
The Entity Relationship Diagram (ERD) models how data flows and interacts:

 Entity 1: User – Attributes: User_ID, Name, Email.

 Entity 2: Biometric_Data – Attributes: Biometric_ID, Biometric_Type, Raw Features.

 Entity 3: Template – Attributes: Template_ID, Cancelable_Temp, Algorithm_Used, Timestamp.

Relationships:
 One User → Many Biometric_Data.

 One Biometric_Data → One Template.

 19

3.9 Assumptions and Dependencies

Assumptions:

 Users will provide clear and valid biometric input (good quality scans).

 Public datasets will be sufficient for training and testing cancelable algorithms.

 Transformation algorithms will maintain an acceptable balance between security and accuracy.

Dependencies:
 Dependent on Python libraries (OpenCV, TensorFlow, NumPy) for processing.

 Dependent on database system (MySQL/SQLite) for secure storage.

 Dependent on hardware scanners for real-world implementation.

If any dependency fails (e.g., dataset unavailable or library compatibility issues), system functionality
may be affected.
 20

3.10 Specific Requirements

The system must meet specific performance and security requirements to ensure reliability:

1. Accuracy: Authentication accuracy should remain above 95%.

2. FAR/FRR Control: FAR should be below 2%, FRR below 5%.

3. Revocability: System must support template reissuance without compromising recognition rates.

4. Non-Invertibility: No attacker should be able to reconstruct raw biometric traits from transformed
templates.

5. Efficiency: Authentication must complete within 2 seconds.

6. Integration: The system should support integration with external applications (e.g., banking,
mobile security).

These requirements ensure that the system is secure, efficient, and scalable for real-world adoption.
 21

References
[1] A. K. Jain, A. Ross, and S. Prabhakar, “An introduction to biometric recognition,” IEEE
Transactions on Circuits and Systems for Video Technology, vol. 14, no. 1, pp. 4–20, Jan. 2004, doi:
10.1109/TCSVT.2003.818349.
[2] N. K. Ratha, J. H. Connell, and R. M. Bolle, “Enhancing security and privacy in biometrics-based
authentication systems,” IBM Systems Journal, vol. 40, no. 3, pp. 614–634, 2001, doi:
10.1147/sj.403.0614.

[3] Z. Zhou, Y. Liu, X. Zhu, S. Zhang, and Z. Liu, “Privacy-preserving cancelable multi-biometric
system for identity information management,” Information Processing & Management, vol. 62, no. 1,
p. 103869, Jan. 2025, doi: 10.1016/[Link].2024.103869.

[4] W. Yang, S. Wang, M. Shahzad, and W. Zhou, “A cancelable biometric authentication system
based on feature-adaptive random projection,” Journal of Information Security and Applications, vol.
58, p. 102704, Jan. 2021, doi: 10.1016/[Link].2020.102704.

[5] J. H. Teng, T. S. Ong, S. M. A. Kalaiarasi, and C. Tee, “Finger vein template protection with
directional Bloom filter,” HighTech and Innovation Journal, vol. 4, no. 2, pp. 412–433, 2023, doi:
10.28991/HIJ-2023-04-02-013.
[6] S. Khan, H. Li, and C. Zhao, “Deep secure PalmNet: A novel cancelable palmprint template
protection scheme with deep attention net and randomized hashing security mechanism,” Computers
& Security, vol. 139, p. 103677, 2024, doi: 10.1016/[Link].2024.103677.

[7] H. Wang, X. Dong, Z. Jin, A. B. J. Teoh, and M. Tistarelli, “Interpretable security analysis of
cancellable biometrics using constrained-optimized similarity-based attack,” in Proc. 2021 IEEE
Winter Conf. on Applications of Computer Vision Workshops (WACVW), 2021, pp. 70–77, doi:
10.1109/WACVW52041.2021.00012.

[8] D. K. Vallabhadas, M. Sandhya, S. D. Reddy, D. Satwika, and G. L. Prashanth, “Biometric

template protection based on a cancelable convolutional neural network over iris and fingerprint,”
Biomedical Signal Processing and Control, vol. 91, p. 106006, May 2024, doi:
10.1016/[Link].2024.106006.

[9]
22
23
24