Group-based ranking method for online rating systems with spamming attacks

Ranking problem has attracted much attention in real systems. How to design a robust ranking method is especially significant for online rating systems under the threat of spamming attacks. By building reputation systems for users, many well-performed ranking methods have been applied to address this issue. In this Letter, we propose a group-based ranking method that evaluates users’ reputations based on their grouping behaviors. More specifically, users are assigned with high reputation scores if they always fall into large rating groups. Results on three real data sets indicate that the present method is more accurate and robust than correlation-based method in the presence of spamming attacks. Introduction. – With the rapid development of the Internet, billions of services and objects are online for us to choose [1]. At the same time, the problem of information overload troubles us every day [2–4]. Therefore, many web sites (Amazon, Ebay, MovieLens, Netlfix, etc.) introduce online rating systems, where users can give discrete ratings to objects. In turn, the ratings of an object serve as a reference and latter affect other users’ decisions [5, 6]. Basically, high ratings can promote the consumption, while low ratings play the opposite role [7]. In real cases, some users may give unreasonable ratings since they are simply unfamiliar with the related field [8], and some others deliberately give biased ratings for various psychosocial reasons [9–13]. These widely existed distort ratings can harm or boost objects’ reputation, mislead others’ judgments, and affect the evolution of rating systems [14–16]. Due to the negative effects of spamming attacks, how to design a robust method for online rating systems is becoming an urgent task [17–19]. To solve this problem, normally, building a reputation system for users is a good way [20–28]. Laureti et al. [25] proposed an iterative refinement (IR) method, where a user’s reputation is inversely proportional to the difference between his ratings and the estimation of the correspond(a)E-mail: [email protected] (b)E-mail: [email protected] ing objects’ quality (i.e., weighted average rating). The reputation and the estimated quality are iteratively calculated until they become stable. An improved IR method is proposed in [26], by assigning trust to each individual rating. Later, Zhou et al. [27] proposed the correlationbased ranking (CR) method that is robust to spamming attacks, where a user’s reputation is iteratively determined by the correlation between his ratings and objects’ estimated quality. Very recently, by introducing a reputation redistribution process and two penalty factors, Liao et al. [28] further improved the CR method. In the majority of previous works [29, 30], a single standard objects’ quality is required in determining users’ reputations, with an underlying assumption that every object is associated with a most objective rating that best reflect its quality. However, in real cases, one object may have multiple valid ratings, since the ratings are subjective and can be affected by users’ background [31–34]. In the presence of more than one reasonable answer to a single task, Tian et al. [29] analyzed the group structure of schools of thought in solving the problem of identifying reliable workers as well as unambiguous tasks in data collection. Specifically, a worker who is consistent with many other workers in most of the tasks is reliable, and a task whose answers form a few tight clusters is easy and clear. Analogously, in the online rating systems, one object’s quality is clear if its ratings are centralized, while it is not clear if