Fake Account Detection in Social Networks with Machine Learning

Details

Supervisor:	David Koll
Duration:	3-6 months
Type:	Master's Thesis (plus Student Project if required)
Status:	open

Fake accounts are a major problem in Online Social Networks such as Facebook, as they can distribute spam, malware or --- more recently --- fake news within the network. Due to the trusting nature of OSN users, these attacks are extremely fruitful from the attacker's perspective. As a consequence, the research community has come up with a vast number of solutions to detect fake accounts (also called Sybil nodes). Our research group has contributed to this research in the past within doctoral and Master's theses [1,2].

The scope of this thesis is two-fold: In a first step, we aim at breaking state-of-the-art solutions in Sybil defenses, and in a second step will target an improvement of their detection capability based on the lessons learned in the first step.

A student taking up this topic should have good knowledge in JAVA or Python (on a level that is sufficient to efficiently implement research like the approach proposed in [3]), basic understanding of classification techniques in machine learning (e.g., Logistic Regression, RandomForest, etc.), evaluation metrics in machine learning (e.g., AUC, False Positives, False Negatives, Precision/Recall) and, most importantly, the ability to critical thinking when faced with research papers (to identify potential attack vectors and weaknesses of the state-of-the-art). The thesis work is likely to result in a peer-reviewed publication, and will be carried out in collaboration with the University of Oregon (USA).

References:

• [1] David Koll, Martin Schwarzmaier, Jun Li, XiangYang Li and Xiaoming Fu: 'Thank You For Being A Friend: An Attacker View on Online-Social-Network-based Sybil Defenses', in Proceedings of IEEE ICDCS 2017 (Workshops): The 9th International Workshop on Hot Topics in Planet-­Scale Mobile Computing and Online Social Networking (HotPOST'17), Atlanta, USA, June 2017
• [2] David Koll, Joshua Stein, Jun Li, and Xiaoming Fu: 'On the State of OSN-based Sybil Defenses', in Proceedings of IEEE/IFIP Networking 2014, Trondheim, Norway, June 2014
• [3] Q. Cao et al: 'Combating Friend Spam Using Social Rejections', in Proceedings of IEEE ICDCS 2015, Columbus, USA, June 2015