AI discovers tricks to bypass internet censorship
Image credit: Dreamstime
A team of researchers has created a machine-learning tool that can ‘learn’ to circumvent internet censorship. So far, it has found dozens of paths around internet censorship in China, India, and Kazakhstan.
Authoritarian governments frequently limit access to information online in order to control narratives and prevent dissent. The 2019 Freedom House report into global internet freedom found that China continued to be the “worst abuser”, while internet freedom declined in the US, Sudan, Brazil, Kazakhstan and many other countries.
Attempts to evade internet censorship have turned into a race to keep up with constantly changing means of limiting access to information online, with internet freedom activists having to manually search for means to evade government censorship. The new tool, named Geneva (short for Genetic Evasion), could give activists an advantage in this race; it trains itself to find workarounds for these restrictions.
Geneva was created by a team of computer scientists led by University of Maryland researchers.
“With Geneva, we are, for the first time, at a major advantage in the censorship arms race,” said Professor Dave Levin, senior author of the study. “Geneva represents the first step toward a whole new arms race in which artificial intelligence systems of censors and evaders compete with one another. Ultimately, winning this race means bringing free speech and open communication to millions of users around the world who currently don’t have them.”
Geneva is a ‘genetic algorithm’: a biologically inspired type of AI which combines small pieces of code (analogous to DNA) to experiment with sophisticated evasion strategies for breaking up, arranging, and sending data packets transferred through the internet. Geneva maintains the small sections of code which are most effective and replaces the rest with each new ‘generation’, refining its ability to evade censorship with each iteration.
A common form of internet censorship works by monitoring these packets sent during searches and blocking requests which contain sensitive keywords (e.g.: Tiananmen Square) or prohibited domain names (e.g.: Wikipedia, Twitter). Using Geneva, these terms could be masked by breaking up and transmitting data in certain configurations, making it much more difficult for censors to restrict access to information.
Through testing against mock censors and against real censors in China, India, and Kazakhstan, Geneva was able to find dozens of previously-undiscovered ways to circumvent censorship through packet manipulation, and also found bugs which the researchers claim are “virtually impossible” find manually.
“This completely inverts how researchers typically approach the problem of censorship,” said Levin. “Ordinarily we identify how a censorship strategy works and then devise strategies to evade it. But now we let Geneva figure out how to evade the censor, and then we learn what censorship strategies are being used by seeing how Geneva defeated them.”
The researchers plan to release their code and findings, with the hope that it will help provide free access to information to citizens of countries with internet censorship. They are also exploring the possibility of implementing Geneva on the server side rather than on the client side, allowing websites like Wikipedia to be accessed by anyone, even those struggling with internet censorship.
Sign up to the E&T News e-mail to get great stories like this delivered to your inbox every day.