Information Assurance: Detection & Response to Web Spam Attacks

Abstract : As online social media applications such as blogs, social bookmarking (folksonomies), and wikis continue to gain its popularity, concerns about the rapid proliferation of Web spam has grown in recent years. These applications enable spammers to submit links that divert unsuspected users to spam Web sites. The goal of this research is to investigate novel techniques to detect Web spam in social media web sites. Specifically, we have developed a co-classification framework that simultaneously detects web spam and the spammers who are responsible for posting them on social media web sites. Using data from two real-world applications, we empirically showed that the proposed co-classification framework is more effective that learning to classify the Web spam and spammers independently. We also investigated an approach to enhance the framework by leveraging out-of-domain data collected from multiple social media web sites.