Mining Latent Semantics from Online Social Media for Cybercrime Forensics

The e-Services provided by HSBC and Hong Kong Stock Exchange underwent cyber-attacks recently. According to a recent cybercrime study performed by Hewlett-Packard, the occurrence of cyber-attacks has doubled over the past three years, and the financial impact of these attacks has increased by 40%. Recent studies have revealed that cybercriminals often communicate and transact through dark markets established in online social media. Such behavior provides an unprecedented opportunity for researchers to tap into online social media for cybercrime forensics, and hence to prevent cyber-attacks. Although data mining on social media have been performed for the analysis of marketing and financial investment activities, its application to cybercrime forensics remains an unexplored area. The proposed research project aims to fill in such a research gap by: (1) developing a novel text mining methodology to extract latent, high-level features (e.g., concepts) from social media to facilitate cybercrime forensics; (2) leveraging the mined high-level features to provide early warnings of various kinds of cyber-attacks before they are launched. To maintain the competitiveness of China's click-and-mortar enterprises, there is a pressing need to develop the proposed methodology to augment existing intrusion prevention systems (IPSs) for cybercrime forensics and prevention.