Emily Wenger: Protecting Privacy and Encryption Standards from AI

Emily Wenger will join the faculty of the Electrical and Computer Engineering Department in Duke University’s Pratt School of Engineering, beginning July 1, 2024. Fresh from completing her PhD at the University of Chicago, Wenger pursues research on two sides of the same AI dice.

In one area of research, Wenger creates programs and filters to protect people’s privacy from AI training algorithms, which scoop up vast quantities of data from all corners of the internet. Through the other concentration, she harnesses the power of machine learning to ensure that the future encryption algorithms that are predicted to stump quantum computer cryptographers are also secure against more traditional cyberattacks.

In both arenas, her work relies on complex mathematics and statistics, which is where Wenger got her start in AI and security. Before earning a master’s degree and working toward her PhD in computer science at UChicago, she earned undergraduate degrees in math and physics from Wheaton College.

She is the recipient of the Federal Graduate Fellowship for STEM Diversity, a Harvey Fellowship from the Mustard Seed Foundation (now 28twelve Foundation), UChicago’s Neubauer and Harper Dissertation fellowships, and a Siebel Scholarship.

“I went to college to study math and discovered, during my post-undergrad work in the US intelligence community, that a lot of issues in the intelligence community and cybersecurity world require interesting work in math, which is how I ended up with two research interests,” Wenger said. “And as we’ve all seen AI evolving as quickly as it has, it’s become clear to me that we need a lot of security and privacy work to keep up with its abilities.”

“As we’ve all seen AI evolving as quickly as it has, it’s become clear to me that we need a lot of security and privacy work to keep up with its abilities.”

Duke Electrical and COmputer Engineering New Faculty member Emily wenger

As quantum computers gain reliability, many organizations and governments across the world are scrambling to build new encryption standards that can stand up to the disruptive technology. With the ability to essentially try every possible solution to break a code at once, quantum computers could theoretically make all previous encryption methods obsolete.

One of Wenger’s ongoing projects relates to a competition run by the National Institute of Standards and Technology to develop a system for protecting data that is not vulnerable to quantum attacks. In collaboration with researchers from Meta AI, she is developing an attack against systems in the NIST competition that uses machine learning to try to recognize connections between encrypted and unencrypted information to discover the underlying methodology.

“These types of attacks are getting stronger all the time,” Wenger said. “Our version hasn’t been able to successfully attack standardized encryption algorithms, but it may well get there in the future.”

Wenger’s second research arm focuses on protecting people’s privacy and creative works from being used to train AI that can recognize faces or recreate artistic styles. With the explosion of recent AI advances such as ChatGPT that can mimic writing prose or paint strokes with the touch of a button, the potential (mis)uses for these types of programs is obvious.

One of Wenger’s projects called Fawkes places a sort of filter or film on top of people’s face images on social media that prevents AI from using them successfully in its training routines. Another, called Glaze, protects artists’ works from a similar fate. Both programs are freely available to the public, with Fawkes alone having been downloaded more than a half-million times.

“These filters act as a stronger version of a watermark that many artists and companies are already using to prevent online theft of their work,” Wenger said. “It’s hard to keep up, because every day there’s a new funky generative AI system that promises to change the world, but maybe it has side effects that haven’t been thought of.”

Wenger also said that she’s excited to join Duke after finishing her work at UChicago. Besides the school’s formidable reputation, she found Duke Engineering to be a very collegial and genuine place when she visited campus. She’s excited by the prospect of collaborating with some notable names in her field, including Neil Gong, who shares her enthusiasm for mixing machine learning with privacy concerns, and Helen Li and Yiran Chen, who direct a handful of large, federally funded centers focused on the implications of AI.

“There’s so many reasons I was excited to get an offer from Duke,” Wenger said. “Besides the school itself, I’m looking forward to making use of the wide variety of camping sites and hiking trails throughout the area.”