Computing User Reputation in Large Online Social Networks by Means of Different Centrality Indices.

Musial-Gabrys, K., Aroyo, L., De Meo, P., Rosaci, D. and Sarne, G.M.L., 2016. Computing User Reputation in Large Online Social Networks by Means of Different Centrality Indices. ACM Transactions on Information Systems. (In Press)

Full text available as:

[img] PDF
trust-networks.pdf - Accepted Version
Restricted to Repository staff only
Available under License Creative Commons Attribution Non-commercial No Derivatives.

1MB

Abstract

In the latest years, a growing number of researchers has been strongly committed to designing algorithms to compute user reputation scores in Online Social Networks (OSNs). Reputation scores can be computed according to two orthogonal perspectives: (a) helpfulness-based reputation (HBR) scores, if users review items available in the OSN and provide feedbacks on the reviews posted by other users and (b) centrality-based reputation (CBR) scores, if the web of trust relationships among users is explicitly available and encoded through a trust network. In HBR approaches, the most reputable users are those who got, on average, the best feedbacks. In contrast, in CBR approaches, the most reputable users are those occupying the most central positions in the trust network and, to this end, multiple indices are now available to compute the centrality of an individual. The goal of this paper is to elucidate the effects of choosing a particular centrality index to compute CBR scores on the robustness of the trust network and to investigate whether CBR and HBR scores are correlated, i.e., that we can predict HBR scores by means of CBR ones. To perform our analysis we used three real-life trust networks extracted from CIAO, Epinions and Wikipedia and we focused on five popular centrality measures: Degree Centrality, Closeness Centrality, Betwenness Centrality, PageRank and Eigenvector Centrality. The main outcome of our study is that, independently of the centrality index we adopt, trust networks gravitate around few highly reputable users whose dismissal seriously injury network robustness. Such a finding has practical implications because the detection of the users having the most destabilizing effect on a trust network allows us for taking suitable counter-measures to preserve network robustness. Furthermore, our study highlights a correlation between HBR and CBR scores and, then, we can design software applications relying on CBR scores to spot users providing the most helpful reviews.

Item Type:Article
Subjects:UNSPECIFIED
Group:Faculty of Science & Technology
ID Code:24644
Deposited By: Unnamed user with email symplectic@symplectic
Deposited On:31 Aug 2016 11:40
Last Modified:31 Aug 2016 11:40

Downloads

Downloads per month over past year

More statistics for this item...
Repository Staff Only -