Shilling recommender systems for fun and profit

Abstract

Recommender systems have emerged in the past several years as an effective way to help people cope with the problem of information overload. One application in which they have become particularly common is in e-commerce, where recommendation of items can often help a customer find what she is interested in and, therefore can help drive sales. Unscrupulous producers in the never-ending quest for market penetration may find it profitable to shill recommender systems by lying to the systems in order to have their products recommended more often than those of their competitors. This paper explores four open questions that may affect the effectiveness of such shilling attacks: which recommender algorithm is being used, whether the application is producing recommendations or predictions, how detectable the attacks are by the operator of the system, and what the properties are of the items being attacked. The questions are explored experimentally on a large data set of movie ratings. Taken together, the results of the paper suggest that new ways must be used to evaluate and detect shilling attacks on recommender systems.

Shilling Attack Detection in Recommender Systems via Selecting Patterns Analysis

“…In 2004, shilling attacks were noticed by researchers who studied the effect of attack models against recommender systems [7]. Since then, numerous efforts have been made to develop methods to detect shilling attacks.…”

“…Unlike normal users, attackers aim at promoting or suppressing the target items [7], [13]. Thus, the selecting patterns of attackers differ from those of normal users.…”

Trust Metrics in Recommender Systems

“…And in fact, gaming standard CF techniques is very easy. While this important aspect has been neglected until recently, recently some recent studies have started to look at attacks of Recommender Systems [17,20]. The simplest attack is the copy-profile attack: the attacker can copy the ratings of target users and fool the system into thinking that the attacker is in fact the most similar user to the target user.…”

“…For example, against shilling attacks [20] in which a user pretends to be similar to the user target of the attack. Trust-aware Recommender Systems can be used to consider only ratings provided by users predicted as trustworthy by the trust metric.…”

A Collusion-Resistant Automation Scheme for Social Moderation Systems

“…In most scenarios, the ratings given by people are subjective and unverifiable; therefore the question Are the raters honest? is an important and challenging problem in this field, as discussed in [3,7,10,11].…”