Feedback behavior and its role in trust assessment for peer-to-peer systems

Abstract

One of the significant characteristics of ubiquitous environments is the need for trust management due to the vital role trust plays in the success of such environments. Recommenders constitute an important component for managing trust because they can corrupt the recommendation network by providing bogus feedback resulting in losing transactions with trustworthy nodes and committing transactions with untrustworthy nodes. Honesty checking schemes, assume that a recommender is consistent and thus does not change its recommendation behavior. In this paper, we shed light on the inconsistent recommenders' behavior. That is, recommenders change their recommendation strategy. To the best of our knowledge, this is the first attempt at linking the behavior change to honesty checking. We propose a new honesty checking algorithm and show through extensive simulation experiments its performance compared to existing honesty checking algorithms.

Comparative trust management with applications: Bayesian approaches emphasis

“…Azzedin and Maheswaran [48] evaluates a trust model for P2P systems that (i) supports multi-level contextual trust, (ii) distinguishes direct/functional and indirect/referral trust, (iii) captures dynamism through temporal decay, and (iv) successfully detects "bad" domains. Azzedin and Ridha [49] investigate "honesty checking schemes" for detecting bogus recommendations and assessing recommenders. This is analogous to detecting badmouthing and ballot-stuffing attacks.…”

A methodology for designing trustworthy e-commerce applications

“…Trust management [21], [22] has been a focus area for research in recent years. A model for supporting trust is proposed in [23].…”

“…For example, researchers in [8] and [9] proposed solutions to mitigate dishonesty. Other researchers [10] and [9] have introduced approaches to address collusion.…”

Research on the implicit feedback mechanism in peer-to-peer trust models

“…For example, in eBay, only 60.7 % of the buyers and 51.7 % of the sellers provided feedbacks to the system after they have received or provided a service [1]. In addition, among the provided feedbacks, there are a lot of false ones, which negatively impact on the correctness of trust computations [2][3][4][5]. Apparently, the existence of false feedbacks and insufficient feedbacks largely reduces a trust model's availability and reliability.…”

“…Apparently, the existence of false feedbacks and insufficient feedbacks largely reduces a trust model's availability and reliability. To tackle the false feedback problem, researchers have proposed the strategies of feedback similarity [2,3] and feedback filtering [4,5], expecting to eliminate the impact of false feedbacks on peers' credibility calculations. Nevertheless, due to the network scale, the technique of feedback similarity suffers from a computational problem, vector sparseness, while feedback filtering has its own problem as well: it succeeds in filtering out false feedbacks, yet some of the true ones are also mistakenly screened out.…”