An Approach to Discover Malicious Online Users in Collaborative Systems
Collaborative filtering systems employ numerous techniques to provide recommendations for online users. These techniques depend on the collected data about online users’ preferences or ratings of provided items, topics, news, etc. Such systems are vulnerable to malicious users’ attacks in which malicious users carefully target specific profiles in order to boost up or diminish the predictions of some targeted items. In this paper, we suggested an approach to find malicious attacks and remove attackers’ profiles. This leads to placing the user ratings in the region of rejection and thereby affecting his level of trustiness, the impact of specific user ratings is affected by the user calculated trustiness level, where completely untrusted user’s ratings will be neglected by recommendation system. We used a Movie Lens of 1 M rating dataset to perform the required training and test the suggested framework. The suggested method distinguished perfectly between Normal, Excess, Inferiority, and completely dishonest users.
- 3.Embarak, O.H., Corne, D.W.: Detecting vicious users in recommendation systems. In: Developments in E-systems Engineering (DeSE), pp. 339–344. IEEE, December 2011Google Scholar
- 4.Panagopoulos, A., Koutrouli, E., Tsalgatidou, A.: Modeling and evaluating a robust feedback-based reputation system for e-commerce platforms. ACM Trans. Web (TWEB) 11(3), 18 (2017)Google Scholar
- 9.Aggarwal, C.C.: Recommender Systems, pp. 1–28. Springer International Publishing, Cham (2016)Google Scholar
- 11.Harper, F.M., Konstan, J.A.: The movielens datasets: History and context. ACM Trans. Interact. Intell. Syst. (TIIS) 5(4), 19 (2016)Google Scholar