Font Size: a A A

Research On The Protection Method Of Face Biometric Template Under The Background Of Deep Learning

Posted on:2021-01-24Degree:MasterType:Thesis
Country:ChinaCandidate:Z Q XuFull Text:PDF
GTID:2438330602995101Subject:Information statistics technology
Abstract/Summary:PDF Full Text Request
Face recognition technology provides a simple,easy and reliable method for human identity recognition,and has been widely used in many scenarios such as face unlocking,face brush payment,access control applications,station and airport security and many other scenarios.Face feature template protection means protecting face features by some method,and protecting the user's face feature information without reducing the face feature template authentication.The face feature template is unique and can't be changed;it will seriously threaten the user's security when the privacy information of the user's face feature template leaked.It means how to ensure the security of the face recognition template becomes particularly important.The accuracy rate of face recognition technology of deep learning has far exceeded the recognition rate of the human eye.As the input of the biometric recognition system,the original face feature template extracted by deep learning needs to be converted into the protected secret template.For the issues of user privacy and security of the original face feature template,this paper proposes a biometric template protection scheme based on deep learning.The original face features are obtained through the deep learning framework and the secret template independent of the face features based on the random projection and random decision tree can be generated,it will give any length of protected face binary feature template vectors,while ensuring the accuracy of face recognition.After experimental simulation and performance analysis,the reversible binary feature template protection scheme proposed in this paper guarantees the accuracy of face recognition,while the protected face template is irreversible and unlinkable,which protects the user Privacy and face feature template security.
Keywords/Search Tags:face recognition, deep learning, biometric template protection, random projection, random decision tree
PDF Full Text Request
Related items