Unorganized Malicious Attacks Detection

Recommender system has attracted much attention during the past decade, and many attack detection algorithms have been developed for better recommendation. Most previous approaches focus on the shilling attacks, where the attack organizer fakes a large number of user profiles by the same strategy to promote or demote an item. In this paper, we study a different attack style: unorganized malicious attacks, where attackers respectively use a small number of user profiles to attack their own target items without any organizer. This attack style occurs in many real applications, yet relevant study remains open. In this paper, we formulate the unorganized malicious attacks detection as a variant of matrix completion problem, and prove that attackers can be detected theoretically. We propose the Unorganized Malicious Attacks detection (UMA) algorithm, which can be viewed as a proximal alternating splitting augmented Lagrangian method. We verify, both theoretically and empirically, the effectiveness of our proposed algorithm.