Shilling Attack Detection Based on Data Tracking

EasyChair Preprint 247

Abstract

Collaborative filtering recommender system is one of the most widely used recommender systems, while it is vulnerable to shilling attack because of its openness. In recent years, many shilling attack detection methods have been proposed and achieved some results. However, with the rapid growth of data, the detection efficiency of existing detection methods can not meet the requirements. To solve the above problem, a detection algorithm based on data tracking adapted to Big Data environment is proposed. Based on two new data features, the algorithm uses extended Kalman filter to track and predict the item's rating, and detects the abnormal item in real-time efficiently. Experimental comparison shows that this algorithm has high detection rate and small time overhead.

Keyphrases: Collaborative filtering recommender system, Data tracking, Extended Kalman Filter, shilling attack detection

@booklet{EasyChair:247,
  author    = {Lingtao Qi and Haiping Huang and Peng Wang and Ruchuan Wang},
  title     = {Shilling Attack Detection Based on Data Tracking},
  doi       = {10.29007/nv9r},
  howpublished = {EasyChair Preprint 247},
  year      = {EasyChair, 2018}}