A Clustering Approach to Unsupervised Attack Detection in Collaborative Recommender Systems

Securing collaborative filtering systems from malicious attack has become an important issue with increasing popularity of recommender Systems. Since recommender systems are entirely based on the input provided by the users or customers, they tend to become highly vulnerable to outside attacks. Prior research has shown that attacks can significantly affect the robustness of the systems. To prevent such attacks, researchers proposed several unsupervised detection mechanisms. While these approaches produce satisfactory results in detecting some well studied attacks, they are not suitable for all types of attacks studied recently. In this paper, we show that the unsupervised clustering can be used effectively for attack detection by computing detection attributes modeled on basic descriptive statistics. We performed extensive experiments and discussed different approaches regarding their performances. Our experimental results showed that attribute-based unsupervised clustering algorithm can detect spam users with a high degree of accuracy and fewer misclassified genuine users regardless of attack strategies.