Sina Kamali and Diogo Barradas. Accepted in NDSS26.
Abstract
Abstract—As Internet censorship grows pervasive, users often rely on covert channels to evade surveillance and access restricted content. Web protocol tunneling tools use websites as proxies, encapsulating covert data within web protocols to blend with legitimate traffic to avoid detection. However, existing tools are prone to detection via traffic analysis, enabling censors to identify the use of such tools via fingerprinting attacks or due to the generation of abnormal browsing patterns.
We present Huma, a new web protocol tunneling tool that addresses existing detection concerns. By deferring covert data transmissions, Huma allows a website participating in circumvention to first respond with unmodified content, while responses embedding covert data are prepared in the background and delivered during the client’s next request, thus avoiding timing anomalies that facilitate fingerprinting. By relying on an overt user simulator modeled after realistic browsing activity, Huma also follows users’ expected browsing behaviors. Lastly, Huma prevents adversary-controlled websites from tying communication endpoints together, enabling straightforward extensions to enable covert communications in Intranet censorship scenarios.
Author’s Note
Huma was one of the hardest projects I’ve ever done. Not only was it another first author project that I was alone on (just me and my supervisor), but this time around, I had to do it in one of the hardest/worst years of my life. To be honest, there were many days that I didn’t even want to complete this project, as I had so many other problems that Huma seemed insignificant compared to them. However, now that I look back at those days, I see that Huma stands as a testament to how I try to get back up again after every time that I’m knocked to the ground.
About the project itself, Huma went through a lot of changes before finally landing on being a censorship circumvention system. The main reason was that I thought the protocol had a lot of potential for many use cases, so I kept pushing the initial design. I am very happy with the final protocol/product, and I deeply hope that it can be integrated with Tor one day as a replacement, or alternative, to WebTunnel.
