Study On The Algorithms Of Privacy Prerving Support Vector Machine

Support Vector Machine (SVM) is a new method of machine learning based on statistical learning theory. SVM has been successfully applied into the fields of auto text classification, handwritten digit recognition, bioinformatics etc. It has many advantages compared to traditional Article Neural Networks or other learning methods, for example, good generalization, being insensitive to high dimension data and convergence to global optimum. It solves the problems of over-learning, dimension curse and local minima etc. In recent years, SVM has become an active research focus in the domain of machine learning. In this paper, based on the problems of privacy preserving arising in the application of traditional SVM algorithms, SVM algorithms of horizontally partitioned and vertically partitioned data were studied. It solves the problem of privacy preserving and has comparable accuracy.SVM is one method of data mining. The concepts, technologies and process of data mining are introduced. Background of SVM, model and algorithms of SVM and the development are presented. The advantages and the applied fields of SVM are summarized in this paper. In the practical applied of SVM, there is the issue of disclosure of personal information. Many researchers pay attention to the issue. In this paper, we study PPSVM of horizontally partitioned and vertically partitioned data.Firstly, we extend the SSP protocol and propose a privacy-preserving algorithm for support vector machine (SVM) classification over vertically partitioned data and this algorithm has better efficiency and security. And we encrypt data using left-multiplied by an invertible square matrix. It also keeps the privacy of the data. Then, we encrypt data using an additively homomorphic encryption and not reveal local kernel matrices of each party. It protects the privacy of the data.