Spam filtering using social networks

Unsolicited commercial email, informally known as spam, is a rapidly growing problem that affects email users. Many methods of filtering spam have been studied; one of them is the use of social networks to distinguish spam from good email. We show that using social networks for filtering spam is practical, effective, and robust. In particular, we investigate one algorithm that builds a graph from a users mailbox, extracts information from the graph, and uses that information to classify email. We show that this algorithm is practical to implement by showing how it can be put into use. We demonstrate that it is effective by giving examples of its use over several datasets, in which it classifies messages with high accuracy and produces no false positives. We show that it can be implemented robustly by identifying several attacks that compromise the filter, and then implementing defenses against these attacks.