In traditional biometrics systems, the system has to store the original biometric features because the biometric feature cannot be reappeared exactly. In permission control system based on biometrics, the biometrics is only used in access control. For example, the fingerprint removable storage devices, the secret files can only be kept in the device and cannot be stored in other devices freely, since the key for encrypt the files has no relationship to the biometric features. Moreover, such technique is easily cracked. In a biometrics password system, the biometrics characteristics of the users are stored in the system directly. Due to the uniqueness of the biometrics, the user of the system cannot change their password while this is a basic property in traditional password system. It also brings the potential privacy issues. While the imposter gets the privacy data, he could get the same right as the user. In traditional cryptography, the security of the key is the most important issue. The information security loopholes are not because of the encryption algorithm, but because of human negligence that makes the key lost or stolen. Thus, in modern information security field, key management has become an important research topic. The biometric key banding method discussed in this paper to solve the three issues above. This method is based on a high accuracy NIR face recognition system. A face key banding method is realized. And the applications based on the method are also discussed. The main contribution of this paper is as follows: 1. An enhanced BioHash method is proposed. This method can convert a vector to a binary string while the binary string can be specified; 2. A face key banding method is realized by combining the enhanced BioHash and BCH error-correcting codes; 3. A multi templates enroll method is proposed. This method can be used for a single template match with multi templates; 4. The applications of this method in face recognition system and other external applications are discussed; 5. A novel linear dimensions reduced algorithm is proposed. This algorithm enhances LDA algorithm via a nonlinear distance weighted function. It is effective for the asymmetric multi-class problem. In general,this paper made a useful explorations research on face encryption algorithms and systems.
修改评论