An Intelligent Recommendation Mobile Application Privacy Risk Evaluation Method Based on Optimized SVM

In recent years, intelligent recommendation (IR) technique has been widely utilized in mobile applications (App), which promotes the development of mobile internet and improves the user experience. However, to provide the service related to users' personalized taste, the mass privacy information of user needs to be collected by IR Apps. This makes IR Apps in front of various security risks, especially the privacy security risk. It is of great significance in both theory and practice to make a privacy risk evaluation on IR Apps. In this work, an approach for IR App privacy risk evaluation based an optimized SVM classifier from a hierarchical privacy security risk factor Set is proposed, which reveals the changing relationship between risk factors and risk levels, and also gives a way to solve the nonlinearity, high-dimensional, and multi-sample problems in the IR App privacy risk evaluation work. In this evaluation model, first we give out a hierarchical privacy security risk factor set oriented the hole data life cycle by comprehensive analysis. Then a feature selection method based on MRMD algorithm is applied to calculate the correlation between features and risk levels, and the factors have a strong correlation with risk level evaluation are selected to constitute a feature vector, which reduce the redundancy. Finally, the extracted features are given input the evaluation model based on an optimized SVM classifier to calculate the risk level. The experiment results show the improvement and a certain generalization in optimized model. This work also enriches the quantitative method in risk evaluation, and can provide quantitative and visual decision support for application providers to improve the privacy protection and for government to keep abreast of development of application security in practice.