Research On Template Security Based On Fingerprint Reconstruction

Fingerprint identification is one of the most widely used biometric identification methods. Most of traditional fingerprint identification systems are based on minutiae, of which the position and direction information is stored in the template database for comparison. With the development of hardware speed and break technology, the whole fingerprint identification system may be entirely in the range of hacker attacks, and the user's identity security and privacy will be threatened. Not like passwords and keys which can be re-set by selecting another one after the old one is lost, the loss of fingerprint feature is permanent. Therefore, the security of fingerprint template becomes our focus and thus a series of template security technology has been developed. Biometric encryption is one of the most typical technologies.This paper concerns the research on the template security. We reconstruct fingerprint image from minutiae template, then attack on an automated fingerprint identification system and a fingerprint encryption system, from which we can draw that fingerprint feature templates do have security risks. Specifically, the main work of this dissertation includes:(1) we implement a FM model based fingerprint reconstruction algorithm to verify the possibility of reconstructing fingerprint images from minutiae, and analyze the experimental results; (2) we implement a matching method which is based on Tico minutia descriptor. This method doesn't rely on the relationship of minutiae and shows strong robustness to the loss of true minutiae and the appearance of spurious minutiae. Based on existing reconstruction and matching algorithms, we design an attack experiment which attacks on a minutiae and orientation descriptor based matching algorithm. Experimental results show that even under a secure setting (0.1% FAR), both type-I attack and type-II attack have a high accept rate, and even the feature used in different systems is not the same, an identification system is still at risk once if the template of another one leaks; (3) Fuzzy Vault method is a typical fingerprint encryption technology. In order to verify that the leak of templates used in traditional identification systems may do bad to existing encryption systems, we introduce a simplified Fuzzy Vault encryption algorithm, then design an attack experiment on an encryption system using reconstructed fingerprint images. Experimental results show that although Fuzzy Vault systems protect minutiae by adding chaff points, once the template of an identification system is compromised, the encryption system on which the same user has registered will be threatened. This is because of the vault concept of Fuzzy Vault that gives attackers chances, and thus reduces the system's security. From the attacker's point of view, this paper has studied security issue about fingerprint identification and encryption systems which are widely used and the shortcoming of security interdependence between systems by experiments. Results in this paper can give help for designing fingerprint identification and encryption systems with high security level.