Detection of Shilling Attacks in Collaborative Recommender Systems

Abstract

Recommender systems are backbone for ecommerce website today. With the rapid growth in the ecommerce web sites, recommendation systems plays very important role to provide personalized service to the user. A good rec ommender system determines the quality of service provided by ecommerce. Ecommerce websites like amazon.com and ebay.com are widely popular be cause of their recommender systems. Collaborative filtering is one of the most widely used recommendation system. Recommendations made using collaborative filtering depend on relationship between the user and items. Unfortunately, due to its openness and dependency on user ratings, Collabo rative filtering is prone to shilling attack and problem is with their security. Such attacks alter the recommendation process to promote or demote a par ticular product. Attacker who cannot be separated with distinguished from genuine user may inject biased profiles in the system to affect the service of system. It may leads to degradation of recommender system’s objective. It is therefore essential to detect the shilling attacks in such a way that there are in-depth analysis of user behaviors and uses two key mechanisms (i.e., behavior features extraction and detection) to distinguish shilling profiles from genuine ones. In the stage of detection, a classifier is then built to dis tinguish attack profiles from genuine user profiles by constructing training data from authentic profiles and attacks generated by attack models. The combined effectiveness of this approach is then evaluated with the supervised classification algorithm Support Vector Machine. The experimental results demonstrate that proposed supervised detection model achieve a better de tection performance is about 85.41% in shilling detection