Cloud-Based User Continuous Authentication And Privacy Preserving

With the rapid development of mobile Internet,the security of mobile devices has become more and more important.At present,most mobile devices require users to perform one-time authentication such as password,fingerprint or face recognition at the time of initial login.However,one-time authentication has a serious security risk.If an authorized user forgets or leaves his authenticated device,the unauthorized user can access the private resource of the authorized user.Continuous authentication can solve the problems of the above one-time authentication.Due to the limited computing and storage resources of mobile devices,many cloud-based continuous authentication systems have been proposed,but these systems are usually designed to pursue higher authentication accuracy without considering the protection of user behavior data privacy.To this end,this paper aims to design a continuous authentication system with high authentication accuracy and privacy preserving.First of all,this paper adopts a continuous authentication system design scheme based on user,service provider and cloud platform,and improves the feature extraction and feature preprocessing modules in the scheme.We use statistical methods to extend features and increase the useful features of the system.In addition,the feature discretization method is used to pre-process the user behavior characteristics,which makes the system's authentication accuracy increase by about 3%.Then,this paper proposes two privacy preserving schemes for the continuous authentication system based on user,service provider and cloud platform.The first is a privacy preserving scheme based on homomorphic encryption,which uses DT-PKC homomorphic encryption algorithm to encrypt user behavior data to protect data privacy.In addition,this paper improves the traditional logistic regression algorithm and designs an PPLR(Privacy Preserving Logistic Regression)algorithm,which can generate an identity authentication model for users based on encrypted data.Compared with the traditional homomorphic encryption privacy preserving continuous authentication system,the scheme has higher authentication accuracy and lower performance overhead.The second is a privacy preserving scheme based on PCA dimension reduction.It uses PCA dimension reduction technology to transform the user's original behavior feature vector into a new user behavior feature vector through spatial transformation,thus achieving the effect of privacy preserving.The scheme uses the One Class SVM algorithm to process user behavior feature vector and generate an identity authentication model for the user.Compared with the privacy preserving scheme based on homomorphic encryption,the PCA dimension-based privacy preserving scheme can greatly reduce the system performance overhead,but its authentication accuracy will be reduced.Finally,this paper implements the above two privacy preserving schemes and constructs multiple experiments to evaluate the system's authentication accuracy and performance overhead.The results show that the two privacy preserving schemes proposed in this paper have higher authentication accuracy than other privacy preserving schemes.Based on the homomorphic encryption privacy preserving scheme,the authentication accuracy rate can reach 93.6%,and it takes 19.2 seconds to authenticate the 127-dimensional user behavior feature vector.Based on the PCA dimension reduction privacy preserving scheme,the authentication accuracy rate can reach 90.8%,and the 127-dimensional user behavior feature vector can be 0.0057 seconds,which greatly reduces the system performance overhead.