THE METHOD OF DETECTING INFORMATION ATTACK OBJECTS IN RECOMMENDATION SYSTEM BASED ON THE ANALYSIS OF RATING TRENDS

Abstract

The subject matter of the research is the process of identifying information attacks on the recommendation system. The goal of this work is to develop a method for detecting information attack objects in a recommendation system based on the analysis of trends in the ratings of system objects. The task to be solved is: develop a method of detecting information attack objects in a recommendation system. Results. The paper investigates methods for determining the existing trends in time series, in particular, methods based on a moving average, several moving averages and zigzag tops. Also, a method for predicting the dynamics of trends in a time series in the future based on R/S analysis was considered. The set of indicators has been proposed, by the values of which it is possible to determine the presence or absence of an information attack on an object of the recommendation system. This set of indicators includes: the existing trend in the numerical series of ratings of the system object, the predicted trend in the ratings of the system object, the number of getting of the object in the lists of recommendations and statistical characteristics of the series, for example, the number of target ratings, the variance of ratings, the variance of the time of assigning ratings, etc. On the basis of the proposed set of indicators, the method for detecting objects of information attack in a recommendation system using trend analysis in the ratings of system objects was developed. This method makes it possible to detect the presence of an information attack on the objects of the recommender system and generates the list of possible targets for bots. Many possible targets can be used to further search for bot profiles and clarify information about their true targets. This will make it possible, when searching for botnets, to check not all system profiles, but only those that interacted with probable targets of attack. Conclusions. The method of detecting information attack objects in recommendation system based on the analysis of rating trends was developed. The software implementation was created and experiments to test the effectiveness of the developed method were carried out. The experiments have shown that the developed method makes it possible to identify with high accuracy the objects of information attacks on recommender systems for random, average and popular attack models.

Keywords

• recommendation systems
• information attacks
• information security
• information attack detection
• technical analysis
• moving average