How games might be the key to avoiding digital censorship

January 7, 2025

How Games Might be the Key to Avoiding Digital Censorship

UMass Amherst computer scientist turns to game theory and AI to study digital surveillance in China and learn how to circumvent the world’s most sophisticated surveillance networks

AMHERST, Mass. – Avoiding digital censorship can sometimes feel like a game of whack-a-mole, with the same success rate, especially when the opponent runs one of the world’s most sophisticated surveillance networks, such as China’s so-called Great Firewall.  But what if out-maneuvering censors was played with a longer view and finer strategy, more like chess?

Amir Houmansadr, associate professor of computer science in the University of Massachusetts Amherst’s Manning College of Information and Computer Sciences, will be extending his previous work on China’s Great Firewall with support from the prestigious Defense Advanced Research Project Agency’s (DARPA) Director’s Award. He will work to devise a comprehensive and secure theoretical foundation involving game theory, artificial intelligence and computational theory, to support far-ranging efforts to support digital freedom.

Avoiding digital censorship tends to go something like this: the censors build a tool to limit what certain users can see or say online. Then those users work to find the inevitable cracks in the censors’ digital tool, and, once they find them, enjoy a degree of digital freedom. At least until the censors discover the workaround, which they’ll address with a new tool, which will soon be circumvented, necessitating another digital tool … and on the game of cat-and-mouse goes.

“What if we could proactively anticipate the tools the censors will use and then develop our circumvention strategies so that we could always stay a few steps ahead,” asks Houmansadr. It’s a very difficult thing to do, almost like becoming a chess grandmaster — and this is where game theory comes in.

In brief, game theory assumes competing interests, each of which wants to “win,” however winning may be defined. In this case, winning would be getting to control the free flow of information. During the course of the “game,” each side has to calculate the potential costs and benefits of each “move,” and to try to project out into the future how those moves will either open up or foreclose different possibilities. “Any action you take has both costs and benefits,” says Houmansadr, “so you’re trying to come up with the best plan to maximize your own benefit. With game theory and some of the recent advances in artificial intelligence, we can lay the foundations for a truly robust theoretically driven strategy.”

Houmansadr, who had previously received $1,000,000 from DARPA’s Young Faculty Award, will first work to build an interactive, game-theory-based framework that leverages a handful of generative AI models that could be used to generate fake web traffic — a sort of distraction to fool censorship.

China has recently started to deploy regional censorship boxes throughout its provinces, and Houmansadr wants to continue his previous work China’s Great Firewall by investigating this emerging censorship strategy.

As the web changes — adopting a protocol known as QUIC — Houmansadr plans to game out how anti-censorship strategy can likewise change in anticipation. And finally, he plans to continue and continuously monitor new forms of censorship as they emerge, and to understand their nature so he can develop the best game-theory-enabled strategy for ongoing circumvention.

“We’re trying to understand something that hasn’t happened yet,” says Houmansadr. “You can’t just go onto the internet and measure it; you need a solid theoretical foundation so that you can successfully avoid censorship even before it develops.”

Contacts: Amir Houmansadr, amir@cs.umass.edu

                 Daegan Miller, drmiller@umass.edu