Issue Details

ENHANCING RECOMMENDER SYSTEM SECURITY: DETECTING INFORMED ATTACKS

ABSTRACT

Several different fields that are relevant to customized services have started using recommender systems. One of the types of it that is used the most often is called a collaborative filtering-based recommender system. However, there is a problem to it, and that is the fact that it is very susceptible to assaults that include profile injection. As a direct consequence of this, the algorithm will now offer biased forecasts. An attacker has a high degree of ease in manipulating the results of these recommender systems due to the poor performance of these systems against these assaults. A number of investigations have been carried out in this field in order to identify these assaults and lessen the damage that they do. Within the scope of this work, we have discussed the informed assaults, which fall under the category of profile injection attacks as well. In order to differentiate these assault profiles from actual users, a variety of distinguishing characteristics have been uncovered. In addition to this, we discussed unsupervised methods for detecting these kinds of assaults with the use of characteristics.

FULL TEXT