Detection of Malicious Attacks in Mobile Operator Traffic Mobile Operator Traffic Recommendation System Based on Graph Convolutional Neural Networks

Abstract

Mobile operators' traffic recommendation system is open to all Internet users, which leads to illegal manipulation of rating data through malicious interference and intentional attacks by unscrupulous users using system design flaws, thus affecting the recommendation results and seriously jeopardizing the security of recommendation services. Most of the existing detection methods are based on manually constructed features extracted from the rating data for TO attack detection, which is difficult to adapt to more complex common access injection attacks, and manually constructed features are time-consuming and lack of differentiation ability, while the scale of the attack behavior is much smaller than that of the normal behavior, which brings unbalanced data problems to the traditional detection methods. Therefore, in this paper, we propose a stacked multilayer graph convolutional neural network to learn the multi-order interaction behavior information between users and items end-to-end to obtain user embeddings and item embeddings, which are used as the attack detection features, and a convolutional neural network is used as the base classifier to realize the deep behavioral feature extraction, and combined with the integrated method to detect the attacks. The experimental results on real datasets show that compared with the popular malicious attack detection methods for recommender systems, the proposed method has a better detection effect on the co-access injection attack and overcomes the problem of unbalanced data to some extent.