报告题目：DeepScan: Exploiting Deep Learning for Malicious Account Detection in Location-Based Social Networks
Yang Chen is an Associate Professor within the School of Computer Science at Fudan University. He received his B.S. and Ph.D. degrees from the Department of Electronic Engineering in Tsinghua University, in 2004 and 2009, respectively. He was a Postdoctoral Associate at the Department of Computer Science in Duke University, and was a Research Associate at the Institute of Computer Science of the University of Goettingen, Germany. His research interests include online social networks, Internet architecture and mobile computing. He is serving as an Editorial Board Member of the Transactions on Emerging Telecommunications Technologies (ETT) and IEEE Access. He served as an OC / TPC Member for several international conferences, including SOSP, WWW, IJCAI, AAAI, IWQoS and ICCCN. He published more than 60 referred papers in international journals and conferences, including IEEE TPDS, IEEE TMC, IEEE TSC, IEEE TNSM, IEEE Communications Magazine, Middleware, INFOCOM, ICDCS, ICDE, CIKM, ACSAC and IWQoS.
The widespread location-based social networks (LBSNs) have immersed into our daily life. As an open platform, LBSNs typically allow all kinds of users to register accounts. Malicious attackers can easily join and post misleading information, often with the intention of influencing the users' decision in urban computing environments. To provide reliable information and improve the experience for legitimate users, we design and implement DeepScan, a malicious account detection system for LBSNs. Different from existing approaches, DeepScan leverages emerging deep learning technologies to learn users' dynamic behavior. In particular, we introduce the long short-term memory (LSTM) neural network to conduct time series analysis of user activities. DeepScan combines newly introduced time series features and a set of conventional features extracted from user activities, and exploits a supervised machine learning-based model for detection. Using the real traces collected from Dianping, a representative LBSN, we demonstrate that DeepScan can achieve an excellent prediction performance with an F1-score of 0.964. We also find that the time series features play a critical role in the detection system.