TOAR: Toward Resisting AS-Level Adversary Correlation Attacks Optimal Anonymous Routing

The Onion Router (Tor), as the most widely used anonymous network, is vulnerable to traffic correlation attacks by powerful passive adversaries, such as Autonomous Systems (AS). AS-level adversaries increase their chances of executing correlation attacks by manipulating the underlying routing, thereby compromising anonymity. Furthermore, these underlying routing detours in the Tor client’s routing inference introduce extra latency. To address this challenge, we propose Toward Resisting AS-level Adversary Correlation Attacks Optimal Anonymous Routing (TOAR). TOAR is a two-stage routing mechanism based on Bayesian optimization within Software Defined Networks (SDN), comprising route search and route forwarding. Specifically, it searches for routes that conform to established policies, avoiding AS that could connect traffic between clients and destinations while maintaining anonymity in the selection of routes that minimize communication costs. To evaluate the anonymity of TOAR, as well as the effectiveness of route searching and the performance of route forwarding, we conduct a detailed analysis and extensive experiments. The analysis and experimental results show that the probability of routing being compromised by correlation attacks is significantly reduced. Compared to classical enumeration-based methods, the success rate of route searching increased by close to 2.5 times, and the forwarding throughput reached 70% of that of the packet transmission. The results show that TOAR effectively improves anonymity while maintaining communication quality, minimizing anonymity loss from AS-level adversaries and reducing high latency from routing detours.