Research On Key Technologies Of The Identity Authentication Basing On Mobile Phone Acceleration Sensor

With the continuous development of informationization and the wide use of smart phone, paying with mobile phone when buying online, e-commerce and online banking also develops. Identity authentication on mobile phone becomes more and more important, the task of which is to identify and verify whether the user visiting the system is legal, real and unique and further decide whether the user can access to such functions as visiting the specific system, resources, service and so on. Identity authentication is the key factor ensuring information security and allocating resources rationally. Basing on analyzing the various current identity authentication methods, the thesis proposes an identity authentication method using mobile phone acceleration sensor dealing with the problems for its application on smart phone.The contribution of the thesis are as follows:1. The thesis designed an identity authentication method on mobile phone for limited resources and configuration, which includes the collection of data, preprocessing of data, extraction of features and pattern matching authentication process. As for the collection of identity authentication data, we collects the acceleration data at every moment when the user uses the mobile phone drawing a track in the space and uses it as an original gesture track with the help of mobile phone acceleration sensor. After that, we preprocesses the data.2. According to the characteristics of signature track acceleration data and combing with wavelet transform theory, the thesis proposes an algorithm using the power of each level after a multi-level decomposition of wavelet as the feature vector of signature track, and applies it into the identity authentication basing on mobile phone acceleration sensor for the first time.3. The thesis comprises and analyses frequently-used model identification methods suitable for identity authentication and points out its shortcomings using on mobile phone, basing on which it chooses support vector machine classification model and uses it to match and verify the feature vector extracted through cross verification to find the optimized parameters to train the model, and finally carries out the whole process of identity authentication method in the thesis.4. We make experiments to evaluate the identity authentication method and the algorithm for feature extraction, compare and analyze the selection of wavelet basis function, the layers of wavelet decomposition, the size of the training model, the original feature, and the selection of C andγin support vector machine. The results tell us that the identity authentication method and the algorithm for feature extraction proposed in this thesis are practical and have the value of promoting. In addition, the thesis has30pictures,6charts and62bibliographies.