Neural Network Detection of Shilling Attack Based on User Rating History and Latent Features

Abstract

The recommendation system is a significant field to solve the problem of information overload today. However, the recommendation system relies heavily on the user's behavior information, which is easy to suffer attacks. Therefore, how to prevent the recommendation system from effecting by the attacks is a major issue in the research field of protection recommendation system. In response to the above questions, we found that the shilling attack has the features of short rating history, large number of rating and other potential features compared with normal user profiles. Therefore, we propose a neural network structure to learn the latent features between users and items and we construct shilling attack detection model based on neural network. The experimental results show that the detection method of this paper has achieved the expected goal.