Research On Network Authentication Based On Face Recognition

Access control is one of the most basic parts of computer network.It is also the key technology of information security and plays an important role in the protection of the system and data security.Biometric identification technology provides a reliable and convenient way to achieve the identity authentication.With the wide use of applications of biometric identification system,an increasing number of researchers pay attention to the security of biometric systems.Face recognition,one of the biometric identification technologies,has become a hot topic for researchers now because of its convenient,friendly and easy to accept.With the development of face recognition,the application of the technology for business use is becoming more and more extensive.The areas of finance which is about everyone's assets like Internet banking,e-commerce are also trying to make use of the face recognition on their authentications.And the security need of such face recognition systems is usually high expected.However,the vulnerability of biometric system to spoof attacks and template attack is also an open problem.Therefore,the thesis focuses mainly on security needs of the face based authentication scheme.Firstly,the thesis introduces the theoretical knowledge of face recognition and analyzes the status quo of the technology.And it is found that the face recognition is vulnerable to the attacks such as facial spoof and template stolen.Based on the vulnerability of face recognition system,we proposed a new framework of face recognition system focused on security.The template protection module and the attack recognition module are added to the framework of the traditional face recognition system,which enhance the anti attack capability and the security of the system.Subsequently,the thesis proposes a new scheme of face template protection based on Fuzzy vault for the added template protection module.The scheme mainly based on the idea of Diffie-Hellman key exchange protocol which is used to transform the face template in the scheme.And we also use the fuzzy vault scheme to store the transformed face template so that the face templates are well protected and the security of the system is enhanced.In addition,the templates used in the proposed scheme could be reset by changing the parameters and help templates if one's template was stolen.Also,the attackers cannot use the template to restore the original biometrics if the template stored in the vault was stolen.Finally,the thesis proposes a novel scheme for facial spoof detection based on analysis of the existing methods of live face detection.Noting the differences of noise included in real human face video and spoof attack video,a method is designed to get the feature of noise frames extracted from the videos.The scheme uses the classifier to judge the real video and attack video which is trained by the extracted features of noise frames.Experiments are simulated on the video database which is taken by us.The results show that the scheme we proposed in this thesis do well in detecting the spoof attack video.