This four-year $300,000 Project Grant from the National Science Foundation's Computer and Information Science and Engineering program aims to develop secure foundations for federated learning. Federated learning enables machine learning models to be collaboratively trained using data from many client devices without sharing private information. The researchers will investigate security vulnerabilities in federated learning's training phase, such as poisoning and backdoor attacks. They will develop provably secure federated learning methods to prevent such attacks and detect malicious clients. The researchers at the University of Massachusetts, as the prime awardee, will incorporate results into courses to train students in developing secure federated learning systems and recruit underrepresented groups. The award supports the NSF's mission to advance computing and information science through investigator-initiated research.
Generated 1/6/24, 3:47 PM