基于时间SFM因子的推荐系统攻击检测方法  

Research of Attack Detection in Recommender Systems in Electronic Commerce

作　　者：唐通[1] 唐雁[2] 沈黎[3] 陈泳序 申红林 王斌 张桉[6] 

机构地区：[1]西南大学数学与统计学院,重庆400715 [2]西南大学计算机与信息科学学院,重庆400715 [3]四川民族学院计算机科学系,四川康定626001 [4]茅台高级中学,贵州仁怀564500 [5]安宁中学,昆明650300 [6]成都七中万达学校,成都610036

出　　处：《西南大学学报（自然科学版）》2013年第11期164-171,共8页Journal of Southwest University(Natural Science Edition)

基　　金：国家自然科学基金资助项目(11101337)

摘　　要：从用户评价行为的时间属性——评价间隔时间入手,提出一种基于时间SFM因子的推荐攻击检测方法,通过仿真实验证明了时间SFM因子的可检测性及其阻止攻击目标项进入用户TopN推荐列表中的有效性.Preference information of users has a significant impact on the recommendation result of the elec- tronic commerce recommendation system and makes it vulnerable to the interference of recommender at tack. Attack detection and attack users＇ discrimination are important for keeping recommender results objective and impartial. An attack detection method based on time SFM factors is put forward in this paper, which represents the characteristics of time gaps from time attributes of user behavior. Simulation experi ment shows that the proposed time SFM factors can be used for testing the recommendation attack and the attack detection method based on time SFM factors in the recommender system can effectively prevent their interferences with TopN recommendation lists for users.

关 键 词：推荐攻击 攻击类型 攻击检测 时间SFM因子 

分 类 号：TP393[自动化与计算机技术—计算机应用技术]