BIOMETRIC TEMPLATE SECURITY

The recent widespread diffusion of biometrics-based recognition systems is mainly due to the greater convenience, comfort and security they offer with respect to traditional authentication methods based on passwords or tokens. In fact, being derived from who a person is or what a person does, instead of from what a person knows or what a person has, biometric data represent identifiers which cannot be lost or forgotten, and represent irrefutable evidences linking a user to his identity.

However, the use of biometric data in an automatic recognition system also involves serious risks for their owners: if a biometrics is somehow stolen or copied, it can be difficult to replace it. Moreover, biometric data can contain sensitive information regarding, for example, the users’ health or genetic background, which can be used in an unauthorized manner for malicious or undesired intents. Moreover the users' privacy can be compromised if cross-matching between different biometric databases is performed, in order to track the enrolled subjects using their personal biometric traits. The aforementioned security and privacy concerns need to be carefully considered when implementing a biometric recognition system, by providing appropriate countermeasures to the possible attacks which can be perpetrated at the vulnerable points of the system. Therefore some measures should be adopted to enhance biometric data resilience against attacks, while allowing the matching to be performed efficiently, thus guaranteeing acceptable recognition performance.

   Among the possible threats regarding users' privacy and security which have to be considered when designing a biometrics-based recognition systems, the unauthorized acquisition of the stored biometric data is probably the most dangerous one. Therefore, many solutions have been investigated in the recent past to secure biometric templates. Among them, cancelable biometrics approaches have been introduced in. These techniques apply intentional non-invertible and repeatable modifications to the original biometric templates. Specifically, a properly defined cancelable biometrics should satisfy the following requirements:

·         Security: it should be impossible or computationally unfeasible to obtain the original biometric template from the transformed one;