Abstract
Despite various mechanisms such as SSL/TLS have been deployed to provide confidentiality and data authentication for web infrastructure, traffic analysis attacks are still able to deteriorate user privacy in web browsing. In this work, we investigate the web page fingerprinting attacks which aim to identify the particular web pages a user has browsed. We find that the traffic occurring while the web browser is rendering a web page exhibits temporal patterns distinguishable by machine learning algorithms. We characterize those patterns as CDN bursts, and use features extracted from CDN bursts to empower classification algorithms to achieve high classification accuracy (96%). 

Bio
Dr. Guangdong Bai is a Senior Lecturer in the University of Queensland, Australia. He obtained his PhD degree from National University of Singapore, and master and bachelor degrees from Peking University. His research interest includes security, software engineering and formal methods. During his previous research, he has worked on analyzing Web protocols, online payment, Android security and IoT security. His research has produced impactful results which contribute to security enhancement of widely-used websites/applications like Sina Weibo, Facebook, Helium, Mozilla and Hadoop, and mobile device vendors like Samsung and Huawei (under CVE and SVE). His work appears on top academic conferences and journals such as NDSS, TSE, ASE and FM.