Evaluating the Effectiveness of Anonymization Systems Against Website Fingerprinting Attacks Using Machine Learning Algorithms: Implementation and Countermeasures
Abdoulwase Al Azzani, Abdulqader Shaawa · 2024
Website Fingerprinting (WF) is a statistical traffic analysis attack, that allows a local, passive eavesdropper to determine a client’s web activity by leveraging features from her packet sequence. These attacks break the privacy expected by users of Privacy Enhancing Technologies (PETs). The basic idea of this kind of attack is to derive a website’s fingerprints from traffic and observed patterns, which can be followed when a website is visited. A library of website fingerprints is created for many interesting websites. This library of previously recorded fingerprints is then compared with the traffic patterns, which could be observed if some anonymization service is used. This lets the attacker learn which websites are downloaded, even if all traffic is encrypted. In this paper, the attack was implemented against the offered protection by a specific anonymization system to reproduce the results reported in the literature. We achieved over 94% accuracy when applied on a new set of monitored sites. Finally, a new suggested costeffective countermeasure was proposed, reducing the success of those attacks to less than 50%.