Cancelable Biometrics with Perfect Secrecy for Correlation-Based Matching

August 26, 2009

Report from Presenter

ICB2009 (International Conference on Biometrics 2009) was held from June 2nd to 5th, 2009 in Alghero in Italy. ICB2009 is a largest international conference for Biometrics, which is organized by IAPR and IEEE. 180 people attended the conference and 130 presentations were made.

At ICB2009, Hitachi, Ltd., Systems Development Laboratory (SDL) made a presentation titled "Cancelable Biometrics with Perfect Secrecy for Correlation-Based Matching," about a new method of cancelable biometrics which has information theoretic security. This presentation partially contains research achievements of a national project funded by Ministry of Internal Affairs and Communications in Japan, "R&D for advancement of functionality and usability in information history management."

Recently, protecting biometric templates has become an issue. It is impossible to revoke biometric unlike password or token, and therefore if biometric is leaked out once and threat of forgery has occurred, the user cannot securely use his biometric anymore. The only remedy is to replace the template with another biometric feature. However, a person has only a limited number of biometric features. Cancelable Biometrics is a biometric verification scheme which was introduced to address this problem. This scheme enables the system to store and match templates while keeping them secret.

In order to realize Cancelable Biometrics, to design the transform function properly is important. First, it is important to preserve the accuracy. Secondly, it is required to prevent the attacker from recovering the original biometric feature from the transformed feature. But none of the existing methods meets both requirements at the same time.

In this presentation, we propose a novel method of Cancelable Biometrics for correlation-based matching (Fig. 1). The biometric image is transformed by Number Theoretic Transform (NTT)^*1, and then the transformed data is masked with a random filter. By applying a particular kind of masking technique, the correlation between the registered image and the input matching image can be computed in masked domain (i.e., encrypted domain) without knowing the original images. It is possible to compute the simple correlation between two images by using NTT. For a matching algorithm based on the simple correlation, we can construct a method of Cancelable Biometrics which multiplies the number theoretic transform of the image by a random filter. And we proved theoretically that in our proposed method the masked version does not leak any information of the original image, in other words, our proposed method has perfect secrecy (Fig. 2). Additionally, we applied our proposed method to finger-vein pattern verification and experimentally obtained very high verification performance.

*1

Number Theoretic Transform (NTT): NTT is Fourier-like transform over a finite field.

(By Shinji Hirata, Systems Development Laboratory)

• page top