Hybrid feature level approach for Multi-biometric Cryptosystem

Abstract

Biometric cryptosystems incorporates the benefits of cryptography i.e. higher flexible security levels and biometric i.e. elimination of tedious task of memorizing the passwords or carrying the tokens. In addition to authentication, this system is used for the task of encryption/decryption or data hiding. The biometric key is generated from a biometric construct using cryptographic key extraction algorithms. Single biometric cryptosystems are seriously affected by various attacks like Intrusion attack, function creep, etc. and hence multiple biometric traits are bound together to form a Multi-biometric Cryptosystem. The threat of breaching the security of the confidential data motivates the development of the data hiding techniques in this paper. The proposed multi-biometric cryptosystem mainly focuses on two phase: 1) multi-biometric fusion and 2) private template protection method. In this paper the fused template of extracted feature is transformed into another domain for more security and mapping is also done in the same domain. It has been observed that equal error rate (EER) of proposed system decreases and the Genuine Acceptance Rate (GAR) of the system increases (GAR 100% with EER 28%) i.e. authentication performance of the system is better as compared to other systems.