Workshop Program

Internet censorship in China is not just limited to the web: the Great Firewall of China prevents thousands of potential Tor users from accessing the network. In this paper, we investigate how the blocking mechanism is implemented, we conjecture how China’s Tor blocking infrastructure is designed and we propose circumvention techniques. Our work bolsters the understanding of China’s censorship capabilities and thus paves the way towards more effective circumvention techniques.

Adversary-resistant communication bootstrapping is a fundamental problem faced by many circumvention (anti-censorship) systems such as Tor. Censoring regimes actively harvest and block published Tor entry points and bridge nodes. More recently, some countries have resorted to reactive (follow-up) probing of the destination hosts of outbound encrypted traffic to identify unpublished Tor nodes. We present the design of a new architecture for bypassing censorship, called DEFIANCE, that extends Tor with resilience to both active harvesting and network scanning attacks. The first goal is accomplished using the DEFIANCE Rendezvous Protocol (RP), and the second is achieved using a novel handshake that we call Address-Change Signaling (ACS). We describe prototype implementations of both components, discuss the limits of our architecture, and evaluate what it would take for a determined adversary to defeat our system. While we develop our prototype components over Tor, their design can be easily extended to other circumvention systems.

User-accessed Virtual Private Network systems allow authorized users remote access to protected or otherwise privileged networks while avoiding dependence on ISPs along the route for data confidentiality and integrity. This direct expression of the internet’s end-to-end principle of security is generally accepted as a highly successful design.

VPN services and technology advertising censorship circumvention, resistance to data retention, and anonymity as features are proliferating rapidly. But it is unclear that these security properties were included in the original design requirements of VPN protocols and product implementations. Experience with dedicated anonymity networks (e.g., Tor) shows that strong anonymity is not achieved by accident. The ‘P’ in VPN notwithstanding, not all privacy methods are equal or strongly anonymizing, which opens opportunities for attackers when VPN-based systems are used for anonymity or even simple censorship circumvention.

This paper evaluates VPN anonymity, security and privacy features including identity, geographic location, confidentiality of communications, and generalized security issues such as reachability and prevention of network tampering. We find many popular VPN products are susceptible to a variety of practical user deanonymization attacks. Weaknesses stem from lack of security analysis of the composition of VPNs, applications, and the TCP/IP stack on each respective operating system. Although we describe some potential mitigations for vendors, the primary goal of this paper is to raise awareness of the inherent risks which come from repurposing off-the-shelf VPN systems to provide strong anonymity.

While mechanics of Web censorship in China are well studied, those of other countries are less understood. Through a combination of personal contacts and Planet-Lab nodes, we conduct experiments to explore the mechanics of Web censorship in 11 countries around the world, including China. Our work provides insights into the diversity of modus operandi of censors around the world and can guide future work on censorship evasion.

The ability to compare two instances of Internet censorship is important because it is the basis for stating what is or is not justified in terms of, for example, international law or human rights. However these comparisons are challenging, even when comparing two instances of the same kind of censorship within the same country.

In this position paper, we use examples of Internet censorship in three different contexts to illustrate the importance of the elements of motivation, resources, and time in Internet censorship. We argue that, while all three of these elements are challenging to measure and analyze, Internet censorship measurement and analysis is incomplete without all three.

The contexts we draw examples from are: public wireless networks in Albuquerque, New Mexico, USA; microblogging in China; and, chat programs in China.

Liberal democracies are increasingly considering internet filtering as a means to assert state control over online information exchanges. A variety of filtering techniques have been implemented in Western states to prevent access to certain content deemed harmful. This development poses a series of democratic and ethical questions, particularly when states introduce regulation mandating ISPs to block online content. In this work we examine the debates surrounding filtering that have played out in two key European states, France and Germany, focusing on the arguments used by opponents and proponents of internet blocking. We use these to explain and analyse the outcomes of both cases and, more broadly, the various challenges posed by internet blocking to democracy.