Detecting Social Spamming on Facebook Platform
Name
Ghada Zakaria Mohamed
Abstract
OSNs (Online Social Networks) are dominating the human interaction nowadays, easing the communication and spreading of news on one hand and providing a global fertile soil to grow all different kinds of social spamming, on the other. Facebook platform, with its 2 billions current active users, is currently on the top of the spammers' targets. Its users are facing different kind of social threats everyday, including malicious links, profanity, hate speech, revenge porn and others. Although many researchers have presented their different techniques to defeat spam on social media, specially on Twitter platform, very few have targeted Facebook's.
To fight the continuously evolving spam techniques, we have to constantly develop and enhance the spam detection methods. This research digs deeper in the Facebook platform, through 10 implemented honeypots, to state the challenges that slow the spam detection process, and ways to overcome it. Using all the given inputs, including the previous techniques tested on other social medias along with observations driven from the honeypots, the final product is a classifier that distinguish the spammer profiles from legitimate ones through data mining and machine learning techniques. To achieve this, the research first overviews the main challenges and limitations that obstruct the spam detection process, and presents the related researches with their results. It then, outlines the implementation steps, from the honeypot construction step, passing through the data collection and preparation and ending by building the classifier itself. Finally, it presents the observations driven from the honeypot and the results from the classifier and validates it against the results from previous researches on different social platforms. The main contribution of this thesis is the end classifier which will be able to distinguish between the legitimate Facebook profiles and the spammer ones. The originality of the research lies in its aim to detect all kind of social spammers, not only the spreading-malware spammers, but also spamming in its general context, e.g. the ones spreading profanity, bulk messages and unapproved contents.
To fight the continuously evolving spam techniques, we have to constantly develop and enhance the spam detection methods. This research digs deeper in the Facebook platform, through 10 implemented honeypots, to state the challenges that slow the spam detection process, and ways to overcome it. Using all the given inputs, including the previous techniques tested on other social medias along with observations driven from the honeypots, the final product is a classifier that distinguish the spammer profiles from legitimate ones through data mining and machine learning techniques. To achieve this, the research first overviews the main challenges and limitations that obstruct the spam detection process, and presents the related researches with their results. It then, outlines the implementation steps, from the honeypot construction step, passing through the data collection and preparation and ending by building the classifier itself. Finally, it presents the observations driven from the honeypot and the results from the classifier and validates it against the results from previous researches on different social platforms. The main contribution of this thesis is the end classifier which will be able to distinguish between the legitimate Facebook profiles and the spammer ones. The originality of the research lies in its aim to detect all kind of social spammers, not only the spreading-malware spammers, but also spamming in its general context, e.g. the ones spreading profanity, bulk messages and unapproved contents.
Graduation Thesis language
English
Graduation Thesis type
Master - Cyber Security
Supervisor(s)
Innar Liiv , Raimundas Matulevičius
Defence year
2018