基于目标项目识别的用户概貌攻击检测算法  

User Profile Attack Detection Algorithm Based on Target Item Identification

作　　者：徐玉辰[1] 梁强[1] 张付志[1] 

机构地区：[1]燕山大学信息科学与工程学院,河北秦皇岛066004

出　　处：《小型微型计算机系统》2011年第7期1370-1374,共5页Journal of Chinese Computer Systems

基　　金：国家"九七三"重点基础研究发展计划项目(2005CB321902)资助;河北省自然科学基金项目(F2008000877)资助

摘　　要：为了准确检测协同过滤推荐系统中的用户概貌注入攻击,从攻击的目的性考虑,通过对攻击概貌中目标项目与填充项目之间的特征差异进行分析,提出一种项目评分背离度的计算方法;利用项目评分背离度来确定受攻击的目标项目,并在此基础上提出一种基于目标项目识别的用户概貌注入攻击检测算法.实验结果表明,该算法进一步提高了攻击检测的精度,确保了系统推荐的质量.In order to detect the user profile injection attacks existing in the collaborative filtering recommender system accurately, we consider the purpose of attack, analyze different characteristics between target item and filler item of attack profiles and propose an approach to compute item rating deviation. Then the attacked target item is determined by exploiting the item rating deviation, on the basis of that, a user profile injection attack detection algorithm based on the target item identification is proposed. The experimental results show that the algorithm improves the precision of detection further, accordingly ensure the quality of the recommendation.

关 键 词：协同过滤 用户概貌注入攻击 攻击检测 项目评分背离度 目标项目 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]