Research Of Shilling Attack Detection Algorithm Based On Characteristic Of Deviate From Time Domain

Recommendation system as a new means of information filtering technology,has achieved good results in dealing with the problem of information overload.However,with the content of the Internet information complexity,the number of access,change and rapid increase of the means of attack,existing recommendation system has exposed many shortcomings.There are some typical example,such as shilling attack.In order to achieve business,illegal businessmen take the probability of deceit to improve their products recommended or lower recommended probability of rival products,resulting in serious decline in the credibility of relevant recommendation information.Therefore,how to ensure the accuracy and security of the recommended information recommendation system has become a problem urgently to be solved at present.This article from shilling attack detection and shilling attack defense research recommendation system security problem in both directions,and reference and research on the current domestic and foreign related literature on the problem of recommendation system supporting attacks.the main work is as follows:Firstly,we choose an anti-attack recommendation algorithm based on score information entropy with better comprehensive performance,validity its effectiveness and convert its result to user trust sorting matrix.In order to reduce the computational complexity of time domain analysis,we apply the user trust sorting matrix to time domain detection algorithm.Secondly,In order to improve the accuracy rate and recall rate of recommendation system detecting shilling attack,a shilling attack detection algorithm based on characteristic of deviate from time domain is proposed.To reduce the complexity of the characteristics of time domain analysis,the user trust sorting matrix is generated by using the anti-attack algorithm that the first step put forward.Then,setting up the positive sequence and the sequence of TOP window,suspicious user sets are generated by intercepting the positive and negative top sequence of user trust sorting matrix.Later,the algorithm extract characteristics of deviate from time domain from suspicious users and combine with gaussian mixture model to realize shilling attack detection.Thirdly,designing experimental scenario based on actual data,the proposed shilling attack detection algorithm in this paper verifys the feasibility and effectiveness of the algorithm.The proposed algorithm has good overall performance.it not only increase attack detection accuracy rate and recalling rate,but also reduce the computational complexity and defense shilling attack.