DEVELOPMENT OF EFFICIENT METHODS FOR BIOMETRIC CRYPTOSYSTEMS

Abstract

Biometric systems make use of physical and behavioural characteristics of individuals for their authentication as these attributes are uniquely associated with them. For ensuring secrecy and authenticity of classified data, Cryptography is used wherein encryption and decryption of data is done using secret cryptographic keys. Thus, a major concern here is to maintain the confidentiality of secret key used for securing information. This issue can effectively be addressed by using Biometric Cryptosystems which combine biometrics and cryptography in order to utilize the best of both domains. Cryptography ensures higher level of security, whereas biometrics provides authentication and non-repudiation. Several biometric cryptosystems based on different modalities have been proposed and developed in the past but these systems suffer from various problems. One major problem is related to unimodal biometric system in which a single biometric characteristic is used for authentication. In such systems, the noise which may creep in during data acquisition process, may lower down the performance of the system. Another issue is related to binding the secret key used in a cryptosystem to biometrics of the user. Third important challenge is protection of the biometric templates stored in the database because if these templates get compromised either through some deterministic method or brute force attack then the user would never be able to use that biometric feature in future. In order to address these issues concerning various aspects of biometric cryptosystems, research work has been carried out and several methodologies have been developed. In order to resolve the first issue, a hybrid multimodal biometric 8 system which combines multimodal features using graph random walk based cross view diffusion, has been proposed. The inherent problem of biometric template protection has also been addressed by transforming each biometric feature value using some pre-defined key features. In addition to this, a novel biometric cryptosystem has been proposed for securing the cryptographic key wherein a secret key is bound with biometric data of the user. New objective functions have been defined for creation of helper data by hiding the secret key. The helper data is subsequently used to retrieve the key. To address the third major problem, an innovative scheme, the Random Area & Perimeter Method (RAPM)) has been proposed wherein biometric characteristics of an individual is transformed into random values that are stored as cancelable biometric templates. Thus, by developing these innovative techniques and methods, all major issues have been addressed for an efficient biometric cryptosystem. The thesis incorporates the developed methodologies, their performance analysis and security analysis along with future directions.