Dr. Swanand Kadhe
Title:FastSecAgg: Scalable Secure Aggregation for Privacy-Preserving Federated Learning
Abstract: In modern large-scale machine learning, federated learning has emerged as an important paradigm, where the training data remains distributed over a large number of clients (e.g., mobile phones, smart devices). In federated learning, each client trains a neural network model locally using their data, and the central server aggregates these local models to obtain an improved model. However, recent attacks have demonstrated that model parameters shared by clients can leak significant amounts of information about their training data, making privacy preservation a critical concern.
In this talk, I will present a secure aggregation protocol, FastSecAgg, that enables the central server to average local models in a privacy-preserving manner while being robust to client dropouts. FastSecAgg reduces the computation cost at the server by several orders-of-magnitude compared to the state-of-the-art schemes, and guarantees security against the server colluding with any subset of some constant fraction (e.g. ~10 %) of clients in the honest-but-curious setting. I will highlight the main building block of FastSecAgg — a novel multi-secret sharing scheme, FastShare, powered by the (finite field) Fast Fourier Transform (FFT). Finally, I will demonstrate that FastSecAgg achieves similar accuracy as vanilla federated averaging on LEAF benchmark datasets for federated learning.
Biography: Swanand Kadhe is a postdoctoral researcher in the EECS Department at the University of California Berkeley. He earned his Ph.D. degree in Electrical and Computer Engineering from Texas A&M University in 2017. He is a recipient of the 2016 Graduate Teaching Fellowship from the College of Engineering at Texas A&M University. He has been a visiting researcher at Nokia Bell Labs, Duke University, and The Chinese University of Hong Kong. From 2009 to 2012, he was an R&D engineer at the TCS Innovation Labs, Bangalore. His research interests lie broadly in federated and distributed machine learning, information and coding theory, privacy and security, and blockchains.