A New Vision for Data Security

In mid-2022, Instagram began asking an assortment of its users to complete a survey about their race, ethnicity and gender. By the end of the year, a huge number of people had filled out the survey, which was part of a new effort to ensure that the social media platform was fair and inclusive.

Behind the scenes, Professor Dawn Song, PhD, an expert in computer security and privacy, was helping ensure that the sensitive data was staying safe while still enabling researchers to analyze it.

“This is the kind of dataset, on hundreds of thousands of people, that you really don’t want to become compromised,” says Song.

Oasis Labs, a startup company that Song launched based on her work at Berkeley, partnered with Instagram’s parent company, Meta to assess the fairness of their artificial intelligence models.  It is the first-of-its-kind, large-scale real-world deployment of advanced privacy computing technologies that Song has been researching in her lab for years. Now, armed with both the success of this large-scale real-world deployment and a grant from the UC Noyce Initiative, Song is working on a second iteration of the platform. Its goal: to make decentralized data science easy.

Rooted in Physics

Song grew up in Dalian, a northern port city in China. In high school in her hometown, she was part of a special class for training students for the Math Olympiad. She then went on to study physics at Tsinghua University. 

“I loved that physics is the language of nature,” says Song. “It describes the world in this elegant, fundamental way.”

But Song had also always been interested in computer science. After one year in a physics PhD program at Cornell University, she decided to switch gears. She completed a master’s degree in computer science at Carnegie Mellon and a Ph.D. at UC Berkeley, focused on computer security and privacy.

“It was a really early time in the field, and the community was very small, but I liked the interdisciplinary nature of the field of computer security,” says Song. “ Security is related to many other fields like economics and policy.”

Song’s research has focused on a kind of catch-22 within computer security: the more locked up and safe digital data is kept, the harder it is to access and use the data. With the advent of machine learning and artificial intelligence— fields that use massive datasets to come to new conclusions— researchers need to develop ways to balance the trade-offs between utilizing data and data privacy.

Taking Back Ownership of Our Data

Song is the co-director of UC Berkeley Center on Responsible Decentralized Intelligence (RDI) and her work on internet and data security revolves around responsible data use and decentralized intelligence. If sensitive information is secret-shared across many different servers, then no one single entity ever has all the information about an individual user.

“With secure multi-party computation, any hacker would need to simultaneously compromise many different entities at once to steal the data,” says Song.

Song is also outspoken about letting internet users maintain better control over their own data. In the future, she imagines people having more say in who uses their data for research and even getting a small payment when they give companies access to information like their demographic data and browsing history. The idea meshes with her security-driven efforts toward decentralized data.

“It’s all about a responsible data economy,” says Song. “How can we help users maintain control of their data and at the same time let the data be used in a privacy-preserving way that is for the betterment of both society and individuals?”

Driven by Real-World Applications

In addition to developing new internet security technologies in her labs, Song has founded or co-founded three start-up tech companies— Oasis Labs, Menlo Security, and Ensighta (acquired by FireEye Inc.)— to move those technologies toward commercial use. Her involvement in the companies, she says, helps her bring more impact of her research to the real world.

“As we give birth to new ideas in research, it’s really nice to be able to see the ideas grow and mature and have real-world impact,” she says.

Her physics background, Song says, helps give her a unique perspective when she tackles computer security problems. And she remains driven by the same fundamental curiosity that first attracted her to science.

“I’m a very curiosity-driven person who loves to explore uncharted domains,” says Song. “I think that’s helped me be a pioneer in my field.”

This article was originally published by UC Berkeley Office of Research.