Table of Contents Author Guidelines Submit a Manuscript
The Scientific World Journal
Volume 2014, Article ID 845897, 8 pages
http://dx.doi.org/10.1155/2014/845897
Research Article

Detection of Abnormal Item Based on Time Intervals for Recommender Systems

1School of Software Engineering, Chongqing University, Chongqing 400044, China
2Key Laboratory of Dependable Service Computing in Cyber Physical Society, Ministry of Education, Chongqing 400044, China
3School of Engineering, University of Portsmouth, Portsmouth PO1 3AH, UK

Received 14 August 2013; Accepted 22 October 2013; Published 12 February 2014

Academic Editors: F. Camastra and J. Tang

Copyright © 2014 Min Gao et al. This is an open access article distributed under the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.

Abstract

With the rapid development of e-business, personalized recommendation has become core competence for enterprises to gain profits and improve customer satisfaction. Although collaborative filtering is the most successful approach for building a recommender system, it suffers from “shilling” attacks. In recent years, the research on shilling attacks has been greatly improved. However, the approaches suffer from serious problem in attack model dependency and high computational cost. To solve the problem, an approach for the detection of abnormal item is proposed in this paper. In the paper, two common features of all attack models are analyzed at first. A revised bottom-up discretized approach is then proposed based on time intervals and the features for the detection. The distributions of ratings in different time intervals are compared to detect anomaly based on the calculation of chi square distribution (). We evaluated our approach on four types of items which are defined according to the life cycles of these items. The experimental results show that the proposed approach achieves a high detection rate with low computational cost when the number of attack profiles is more than 15. It improves the efficiency in shilling attacks detection by narrowing down the suspicious users.