時間: 2025/12/15 上午10:30

地點: 大仁樓200301教室

內容摘要:

Federated learning (FL), which enables Artificial Intelligence of Things (AIoT) devices to collaboratively train shared machine learning (ML) models without exposing their raw data, has emerged as a promising paradigm for distributed ML while preserving data privacy among AIoT devices. To facilitate FL in wireless networks, the over-the-air (OTA) computation, which leverages the waveform-superposition property of wireless signals to simultaneously perform communication and computing tasks over wireless multiple-access channels, has recently attracted growing research attention. By integrating the OTA computation into FL, locally trained models can be aggregated wirelessly without incurring excessive bandwidth consumption. In this talk, we introduce a cohort-based power scaling and gradient recovery algorithm specifically designed for OTA-FL systems. Recognizing that OTA-FL systems are inherently vulnerable to jamming attacks due to the difficulty of decoupling gradient and jamming signals, we investigate strategies to safeguard OTA-FL systems against such attacks. In addition, since the gradient signals in OTA-FL systems may be opportunistically intercepted by passive wardens, we further examine mechanisms that provide covertness to prevent these signals from being eavesdropped.

講者簡介

Dr. Yi-Han Chiang is an Associate Professor in the Department of Electrical and Electronic Systems Engineering, Osaka Metropolitan University, Osaka, Japan. He received the Ph.D. degree from the Graduate Institute of Communication Engineering, National Taiwan University, Taipei, Taiwan, in 2017. He was a Postdoctoral Researcher with the Information Systems Architecture Science Research Division, National Institute of Informatics, Tokyo, Japan, in 2018 and 2019. He served as an Assistant Professor in the Department of Electrical and Information Systems Engineering, Osaka Prefecture University, Osaka, Japan, from 2020 to 2022. His research interests include wireless federated learning, age of information, mobile edge computing, and energy efficiency of wireless networks.