An effective recommender attack detection method based on time SFM factors

Users Preference information has significant impact on the recommendations. It makes recommender system vulnerable. To make detection and discrimination of attack users accurate and recommendations objective, time intervals of user's rates was taken into consideration. After a series of Rate-time pretreatment, SFM factors short for span, frequency and Mount properties were summed up, representing time attributes of user behaviors. An effective attack detection method based on time SFM factors is proposed to more effectively prevent their interferences with TopN recommendation lists for users. Experiment results support the conclusion.