An Ensemble Method For Detecting Shilling Attacks Based On Ordered Item Sequences

With the rapid development of the network, people feel the unprecedented convenience, people can never leave home shopping, communication, leisure and entertainment, etc. But the Internet bring side effects are also obvious, when people are faced with huge amounts of information are often unprepared, difficult to distinguish authenticity, have no choice, we call this phenomenon "information overload". Collaborative filtering recommendation system well solved the problem of "information overload", and widely used, its application fields include electronic commerce, multimedia, etc. The caveat is collaborative filtering recommendation system is not perfect, due to defects of the recommendation system itself some attackers for commercial interests to recommend against a wide variety of goods, to the end user to make false recommendation, which ultimately leads to change the initial attitude toward the trust of the recommendation system. In order to solve this problem, scholars have developed many attack detection methods for recommendation systems, the main problems of these methods is to reduce the false alarm rate and improve the accuracy. Collaborative recommender systems are vulnerable to shilling attacks in which malicious users deliberately manipulate their recommendation output by inserting fake profiles into the systems. To reduce this risk, a number of methods have been proposed to detect such attacks. However, most of them suffer from low precision.To solve this problem, we propose an ensemble method to detect shilling attacks based on ordered item sequences(EMDSA-OIS).Firstly, by analyzing the differences of rating patterns between genuine and attack profiles, we construct ordered popular item sequences and ordered novelty item sequences, and based on which the popular and novelty item rating series are constructed for each user profile.Secondly, we propose six features to characterize the attack profiles. Particularly, we extract two features based on the popular and novelty item rating series. We partition item set according to the ordered item sequences, and combine with mutual information to extract another four features.Finally, we propose an ensemble framework to detect shilling attacks. In particular, we create base training sets with great diversities using bootstrap resampling technique. Based on these base training sets, we train decision tree algorithm to generate diverse base classifiers. The simple majority voting strategy is used to combine the predictive results of these base classifiers. Experimental results show that the proposed method can effectively improve the precision while keeping a high recall.