A CNN-based Hybrid Model and Architecture for Shilling Attack Detection

Recommendation systems are widely used in various areas to personalize recommendations and suggestions to users. However, they are vulnerable to shilling attacks in which malicious users try to promote their products or diminish their competitors'. Therefore, detecting shilling attacks can significantly improve the quality of recommender systems and user experience. With the increasing complexity of attacks and changes in attackers' behavior, more advanced approaches are required to find the hidden patterns in data. This paper proposes a CNN-based hybrid model and architecture to integrate self-learning and flexible aspects of CNN with other approaches to enhance the prediction results of shilling attacks. Two benchmark datasets are used in the experimental analysis, the Movie-Lens 100K and Netflix. The performance of the proposed hybrid models is compared to that of the traditional deep learning and machine learning detection methods. Experimental results show that the superiority of hybrid models over individual models depends on the sparsity level of data and divergence of results.