A Face Feature Encryption Algorithm Based On Fuzzy Extraction

Biometric cryptosystems refer to biometric authentication systems that utilizetechniques of binding biometrics and information security systems togetherseamlessly. Characteristics of the human face have been widely used with theperformance of universality, intuition and non-invasion, combining it with passwordcan overcome the shortcomings that traditional face recognition systems maydisclose privacy and guarantee people’s needs for information safety. However, atthis stage, the research in this area is poor, so it is necessary to do in-depth study onthe facial feature encryption, and following the world’s advanced level.The fuzzy extractor algorithm is researched which can generate accurate anduniform distribution of key, and proposes the encryption scheme of the facial featurebased on fuzzy extraction, then improves the method of storing data in the database,and effectively improving the security of the system.Firstly, traditional cryptography requires that keys must be, but biologicalcharacteristics can’t obtain stable and precise key for the presence of noise andinterference of other factors. To resolve this problem, the system obtain the publicinformation and the private information from the initial biometric image with fuzzyextractor algorithm, when second obtained biometric image has a small changecompared to initial image, this algorithm uses the first generated public informationand the second obtained feature image to restore a secret key, which complies withthe requirements of the cryptography.Secondly, a facial feature encryption scheme based on the fuzzy extractor isproposed, which include two stages of registration and recognition: in the period ofregistration, extract the feature information from input face image, then form out theintermediate information of uniform distribution by the method of Fault-tolerant. Atlast, get the public information and the private information from the intermediate information by Encoding and Hash function; in the period of recognition, use thepublic information and the new feature information to decoding, recover the privateinformation by Hash transformation, when the result of matching is same then theauthentication is successful. This scheme can stably extract the face featureinformation, and store the encrypted feature information instead of the originaltemplate in the database, so it effectively protect the security of biometricinformation and user’s privacy, and applicable to the protection of the face template.Finally, the system constructs the facial feature encryption system based onfuzzy extractor in the Visual C++6.0development environment, and tests theperformance of the encryption system in the ORL face database of CambridgeUniversity and self-built face database. Experimental results show that: in the ORLlibrary which has400face images3face image registration, recognition rate reaches87%, false acceptance rate is0.25%, and the performance index can meet therequirements of practical application.