A potential HTTP-based application-level attack against Tor

Abstract

Tor has become one of the most popular overlay networks for anonymizing TCP traffic, however, the anonymity of Tor clients is threatened by various attacks exploiting traffic analysis or Tor’s design features. Although considerable effort has been made to secure and improve Tor networks, little attention has been paid to various application-level attacks against Tor. In this paper, we present a potential HTTP-based application-level attack against Tor, which exploits both Tor’s design features and HTTP’s vulnerability to man-in-the-middle attacks. Such an application-level attack can efficiently and effectively compromise the anonymity of clients without using invasive plugins like Java or any other active content systems in a web browser, posing a serious threat to Tor. Our analytical and empirical results validate the feasibility and effectiveness of the attack. Based on our analysis of the potential attack mechanism, we propose corresponding countermeasures to thwart such potential application-level attacks against Tor, thereby effectively securing and improving Tor networks. Since the fundamental vulnerability exposed by this paper is not specific to web browsing via Tor but rather to the problem of other low-latency applications based on TCP streams, our study is critical for securing and improving low-latency anonymous communication systems.