Website Fingerprinting Attacks with Advanced Features on Tor Networks

The Tor network has been discovered to be susceptible to website fingerprinting (WF) attacks. Previous research has primarily been conducted in controlled experimental environments, leading to debates about the feasibility of WF attacks in real-world environments. Recent advancements in feature engineering and machine learning models have aimed to bridge this gap by exploring real-world scenarios. Nonetheless, designing innovative features for WF attacks on Tor networks remains a significant challenge. To address these challenges, this research explores real-world environments using a novel method that extracts network traffic data by filtering out common traffic based on onion service protocols and advanced features that optimize various advantages of different features. The results suggest that WF attacks in real-world environments become more feasible with the proposed method and carefully crafted features. This study contributes to our understanding of the feasibility of WF attacks on Tor networks in real-world environments and can help identify potential security enhancements on Tor networks.