Unsupervised Spam Detection in Hyves Using SALSA.

With the escalation in popularity of social networking sites such as Twitter, Facebook, LinkedIn, MySpace, Google+, Weibo, and Hyves, the rate of spammers and unsolicited messages has increased significantly. Spamming agents can be automated spam bots or users. The main objective of this paper is to propose an unsupervised approach to detect spam content messages. In this paper, stochastic approach for link-structure analysis (SALSA) algorithm is used to classify a message being spam or not-spam. The dataset from the popular Dutch social networking site named Hyves has been obtained and tested with different performance measures namely true positive rate, false positive rate, accuracy, and time of execution, and it is found that this mechanism outperforms the previously existing unsupervised author-reporter model for spam detection based on HITS.