An Unsupervised Approach for Identifying Spammers in Social Networks

This paper proposes an unsupervised method for automatic identification of spammers in a social network. In our approach, we first investigate the link structure of the network in order to derive a legitimacy score for each node. Then we model these scores as a mixture of beta distributions. The number of components in the mixture is determined by the integrated classification likelihood Bayesian information criterion, while the parameters of each component are estimated using the expectation-maximization algorithm. This method allows us to automatically discriminate between spam senders and legitimate users. Experimental results show the suitability of the proposed approach and compare its performance to that of a previous fully-supervised method. We also illustrate our approach through a test application to Yahoo! Answers, a large question-answering web service that is particularly rich in the amount and types of content and social interactions represented.