好色tv

报告人： Xintao Wu University of North Carolina at Charlotte

题 目： Privacy Preservation and Fraud Detection in Online Social Network Analysis

时 间：2014年06月19日 10:00

地 点：九号教学楼北307B

报告人简介:

Dr. Xintao Wu is a Professor in the Department of Software and Information Systems and director of Data Privacy Lab at the University of North Carolina at Charlotte, USA. He will join University of Arkansas this August as a professor and the Charles D. Morgan/Acxiom Endowed Graduate Research Chair in Database. He got his Ph.D. degree in Information Technology from George Mason University in 2001. He received his BS degree in Information Science from the University of Science and Technology of China in 1994, an ME degree in Computer Engineering from the Chinese Academy of Space Technology in 1997. His major research interests include data mining, data privacy and security, and social network analysis. His recent research work has been to apply spectral analysis for fraud detection in social networks and develop privacy preserving data mining techniques for linked data in social networks. Dr. Wu is an editor of Journal of Intelligent Information Systems, Transaction on Data Privacy, International Journal of Social Network Mining, and a guest editor for a special issue of IEEE/ACM Transactions on Computational Biology and Bioinformatics. He frequently served on program committees of top international conferences, including ACM KDD, CIKM, IEEE ICDM, SIAM SDM, PKDD, and PAKDD, and NSF review panels. Dr. Wu is a recipient of NSF Career Award in 2006 and college's Outstanding Faculty Research Award from UNC Charlotte in 2009. He and his students received several best paper awards including PAKDD'09 Best Student Paper Runner-up Award, WISE'12 Challenge Runner-up Award, PAKDD'13 Best Application Paper Award, and BIBM'13 Best Paper Award.

报告提纲：

Social networks have received dramatic interest in research and development. In this talk, We will focus on two major challenges in online social network analysis: privacy preservation and fraud detection. Social networks often contain some private attribute information about individuals as well as their sensitive relationships. The privacy concerns associated with data analysis over social networks have incurred the recent research on privacy-preserving social network analysis, particularly on privacy-preserving publishing, querying and mining social network data. Compared with well studied anonymization and perturbation techniques for tabular data, it is more challenging to design effective privacy preservation techniques for publishing and mining social network data because of the difficulties in modeling link structures in social networks. In the first part of this talk, we present our recent research on both randomization based privacy preserving social network data publishing and differential privacy preserving social network mining based on output perturbation. Social networks are also vulnerable to both large-scale attacks (e.g., spam, denial of services, Sybil attacks) and subtle anomalies. Traditional topology-based detection methods often fail to effectively identify those collaborative attacks and subtle anomalies. In the second part of this talk, we present a spectral analysis based fraud detection framework in which we examine the eigenvectors of the adjacency matrix of the underlying graph topology. We conduct theoretical analysis to show attacking nodes locate in a different region of the spectral space from regular nodes. In particular, we focus on 1) Random Link Attacks in which the malicious user creates multiple false identities and interactions among those identities to later proceed to attack a large number of randomly chosen users of the network; and 2) subtle anomalies that are often embedded within a community but are structurally dissimilar to the background. In the summary portion of this talk, we will discuss the applicability of our techniques to other domains including healthcare, e-commerce, and genome wide association studies. Sample projects will be briefly presented.