Princeton electrical engineering researchers have received funding from Facebook to develop an AI model that protects private data while allowing technology companies to perform aggregate data analytics.
The team, headed by Associate Professor of Electrical Engineering Prateek Mittal and led by graduate student Sameer Wagh, proposed a three-party model that boosts the efficiency and the security of deep neural networks requiring the use of private data. The approach is especially important for policing malicious actors on social media.
The team was one of 10 award winners in Facebook's request for proposals in "systems and machine learning," chosen from a pool of 167 proposals from more than 100 universities across 26 countries. In a news release announcing the award, Kim Hazelwood, senior engineering manager at Facebook, expressed her excitement for the "investments in academic research in this important domain."
In addition to analyzing online platforms, the researchers said their method could improve medical technologies used to identify and treat rare diseases, linking data across hospitals while maintaining secure protocols.
The team received $50,000 for their proposal, titled "Efficient and Private Deep Learning using 3-party Secure Computation."