Face Template Security
Security and privacy concern is one of the most important issues in biometric recognition systems. Since there are intra-class variations existing in biometric templates, the traditional encryption methods which are sensitive to variations are not available to protect biometric data. Our project goal is to develop reliable face template protecting schemes, which has high security, reliable accuracy and cancelability.
The right figure explains the main purpose of our project. The above block diagram shows the structure of the biometric system. We mainly want to protect the templates stored in database. And the below one describes the procedure to implement this protection.
1. The Class-Distribution-Preserving Transform [2,3]
There are already cryptosystem schemes proposed to enhance the security of biometric templates. Unfortunately, most of these schemes require binary templates as input. Our proposed "CDP transform" scheme transforms the original face templates into binary strings. It selects a series of distinguishing points (B), computes the distances between the original templates and the distinguishing points. The distances are thresholded to get binary strings as the final binary representation.
2. The Hybrid Framework 
We expand the CDP transform scheme to construct a three-step hybrid framework, such that the constructed scheme can achieve the three requirements: cancelability, discriminability and security by integrating three different schemes together. Each one of these three schemes provides one property respectively.
Experimental Results with The Hybrid Framework: (three different databases: CMU PIE, FERET, FRGC)
Histogram: The three figures show the histograms of the feature templates with three different variations (pose, illumination, pose & illumination) with the CMU PIE database. The subfigures (a), (b) and (c) show the genuine and imposter distribution of the feature templates in each step. The distribution of the binary templates has the smallest overlapping rate. Subfigure (d) shows the performance of the feature templates in each step.
Performance: Show the performance with the three databases.
11/10/2009. PPT for ACM-MM-ICME 2009.
2. Y C Feng and P C Yuen, "Selection of Distinguish Points for Class Distribution Preserving Transform for Biometric Template Protection," Proceedings of IEEE International Conference on Biometrics (ICB), pp. 636-645, 2007.
3. Y C Feng and P C Yuen, "Class-Distribution Preserving Transform for Face Biometric Data Security," Proceedings of IEEE International Conference on Acoustics, Speech, and Signal Processing (ICASSP), pp. 141-144, 2007.
Attacks on Biometric Systems:
C. Hill, “Risk of Masquerade Arising from the Storage of Biometrics,” B.S. Thesis, Australian National University, http://chris.fornax.net/biometrics.html.
M. Martinez-Diaz J. Fierrez-Aguilar F. Alonso-Fernandez J. Ortega-Garcia J.A. Siguenza, “Hill-Climbing and Brute-Force Attacks on Biometric Systems: A Case Study in Match-on-Card Fingerprint Verification,” Proceedings 40th Annual IEEE International Carnahan Conferences Security Technology, 2006
Survey on Biometric Protection Schemes:
Fuzzy Vault Scheme:
S Yang and I Verbauwhede, “Automatic secure fingerprint verification system based on fuzzy vault scheme,” in Proceedings of the IEEE International Conference on Acoustics, Speech and Signal Processing, vol. 5, pp. 609-612, March 2005.
S Tulyakov, V Chavan, and V Govindaraju, “Symmetric Hash Functions for Fingerprint Minutiae,” Proceedings of the Int’l Workshop Pattern Recognition for Crime Prevention, Security, and Surveillance, pp. 30-38, 2005.
A Teoh, A Goh and D Ngo, “Random Multispace Quantization as an Analytic Mechanism for BioHashing of Biometric and Random Identity Inputs,” IEEE Transactions on Pattern Analysis and Machine Intelligence, vol. 28, no. 12, pp. 1892-1901, 2006.