联邦学习学术报告|Federated Learning: Privacy, Efficiency, and Robustness
原文来自公众号：浙大网安
链接：https://zoom.us/j/9828106847
Federated Learning: Privacy, Efficiency, and Robustness
报告摘要：
Federated learning (FL) has received increasing attention in both industry and academia by enabling multiple clients (e.g., mobile devices or institutions) to jointly train machine learning models while keeping data at local clients. There are several well-recognized challenges including: 1) privacy: while FL avoids direct exchange of the local data, the model updates being exchanged could still be used to infer sensitive information of the local data, 2) communication efficiency: the multi-round high-dimensional model updates often incur significant communication overhead, 3) robustness: the training is vulnerable to Byzantine failures and adversarial attacks of the clients such as data poisoning and label flipping attacks.
In this talk, I will present several of our recent works addressing these challenges including: 1) Projected Federated Averaging (PFA), which optimizes model utility while ensuring formal differential privacy of the model updates given heterogeneous privacy requirements of clients and minimizing communication cost (VLDB ’22), 2) Federated Pruning, which trains a reduced model to reduce communication overhead while maintaining similar performance compared to the full model (INTERSPEECH ’22), 3) Robust Aggregation (RobustFed), a truth inference approach inspired from crowdsourcing for robust federated learning that learns and incorporates clients’ reliability into model aggregation (CIKM ’22). I will conclude by discussing open directions that explore the synergy among these three challenges.
报告人简介：
Li Xiong
-ACM Distinguished Member
-IEEE Fellow-埃默里大学教授Li Xiong is a Professor of Computer Science and Biomedical Informatics at Emory University. She held a Winship Distinguished Research Professorship from 2015-2018. She has a Ph.D. from Georgia Institute of Technology, an MS from Johns Hopkins University, and a BS from the University of Science and Technology of China. She and her research lab, Assured Information Management and Sharing (AIMS), conduct research on the intersection of data management, machine learning, and data privacy and security. She has published over 170 papers and received six best paper or runner up awards. She has served and serves as associate editor for IEEE TKDE, VLDBJ, IEEE TDSC, general or program co-chairs for ACM CIKM 2022, IEEE BigData 2020, and ACM SIGSPATIAL 2018, 2020. Her research has been supported by National Science Foundation (NSF), AFOSR (Air Force Office of Scientific Research), National Institute of Health (NIH), and Patient-Centered Outcomes Research Institute (PCORI). She is also a recipient of Google Research Award, IBM Smarter Healthcare Faculty Innovation Award, Cisco Research Awards, AT&T Research Gift, and Woodrow Wilson Career Enhancement Fellowship. She is an IEEE fellow. More details are at http://www. cs.emory.edu/~lxiong.
往期推荐
隐私计算机器学习中，应用MPC实验常见问题与解答
Piranha:用于安全计算的GPU平台阿里、浙大顶会论文：联邦环境下，基于元学习的图谱知识外推本地差分隐私 VS 全局差分隐私