OUStralopithecus

Abstract

In many parts of the world, censors are continuously increasing their capacity to fingerprint, identify, and block censorship resistance tools to maintain control over what can and can not be accessed over the Internet. In response, traffic replacement, which involves co-opting a steady stream of uncensored overt traffic to serve as a perfect cover for censored covert content, has been developed in an effort to provide undetectable access to the open Internet for those in censored regions. Despite the promise of this technique, creating a suitable stream of uncensored overt traffic that is high throughput, fingerprint and identification resistant, and does not overburden the user to generate, is an underexplored area that is critical to traffic replacement's success. To address this, we propose OUStralopithecus (OUStral for short), a web-based Overt User Simulator (OUS) that browses the web as a human would in order to avoid being detected by a censor. We implement OUStral as a Python library that can be added to an existing traffic-replacement system. To evaluate OUStral we connect it to an existing traffic replacement system, Slitheen, that replaces media data such as images. Additionally, we implement WebM video replacement for Slitheen to demonstrate the high throughput that OUStral is able to provide. We show that OUStral evades being detected as a bot by state-of-the-art bot detection software while providing a high-throughput overt data channel for covert data replacement.