Neil Gong: At the Intersection of Security, Privacy and Machine Learning

Neil Gong joined the Department of Electrical and Computer Engineering in the Duke University Pratt School of Engineering on July 1, 2019. An expert in digital security technologies, Gong is one of a handful of researchers at the forefront of exploring privacy and security issues and techniques related to machine learning and artificial intelligence.

“Machine learning is being deployed in many aspects of our societies like self-driving cars, precision healthcare and cybersecurity,” said Gong, who joins Duke’s faculty from Iowa State University. “And hackers are always motivated to follow emerging technologies to find new vulnerabilities as well as new tools for their trade.”

Gong earned a B.E. in computer science from the University of Science and Technology of China in 2010 before completing his PhD in the same field at the University of California – Berkeley in 2015. It was there Gong got hooked on the subject.

“Now that machine learning applications are being widely used, the intersection between privacy and security is an emerging area of research. Stanford, MIT, Berkeley, CMU and Google all have groups focused on these questions. I decided to come to Duke and stay in academia because of the flexibility it provides in answering more fundamental research questions that might take years to pay off commercially.”

Gong’s research focuses on two aspects of the intersection of privacy, security and machine learning. The first is trying to figure out how malicious actors might misuse machine learning techniques to attack computer and network systems as well as how to leverage machine learning to enhance cybersecurity and privacy. The second is to study the vulnerabilities of machine learning itself and how to build secure and privacy-preserving machine learning.

For example, when large tech companies such as Facebook or Google implement machine learning algorithms, they are typically embedded within the programs or applications themselves. This leaves them vulnerable to hackers to design their own machine learning algorithms that can eventually reverse engineer the intellectual property being used.

Another example in the area of privacy is how some companies such as Cambridge Analytica are working to mine data from user profiles across social media platforms. While some of this activity is legal, other efforts use machine learning to try to use what data is available to infer private information purposefully left out of public profiles.

“But it might be possible to defend against attacks like these by including a few pieces of false information in a profile,” explained Gong. “The question is how many pieces are needed and what types of information work best.”

Another aspect of Gong’s research deals with finding ways of keeping medical data secure even when it’s being used by large machine learning research projects designed to improve patient health. This line of study could find wide use for Duke’s emerging leadership in this field through programs such as Duke Forge.

“Now that machine learning applications are being widely used, the intersection between privacy and security is an emerging area of research,” said Gong. “Stanford, MIT, Berkeley, CMU and Google all have groups focused on these questions. I decided to come to Duke and stay in academia because of the flexibility it provides in answering more fundamental research questions that might take years to pay off commercially.”