Toward Low-Latency Privacy-Preserving Federated Learning in Fluid Antenna Systems with Homomorphic Encryption

  • Ho, Manh Cuong; 
  • Dinh, Ton That Tam; 
  • Tran, Kiet Nguyen Tuan; 
  • Mac, Huy Dang; 
  • Oh, Junsuk; 
  • ... Cho, Sungrae; 
  • 외 2명
Citations

SCOPUS

0

초록

This paper investigates federated learning-assisted fluid antenna systems (FL-FAS), focusing on the joint optimization of fluid antenna (FA) positioning and ciphertext expansion factor to minimize uplink transmission time with homomorphic encryption (HE). Federated learning (FL) has been widely adopted as a robust, secure and privacy-preserving paradigm for wireless communication networks, enabling distributed devices to collaboratively train models without sharing raw user data, while still contributing effectively to the construction of a global intelligence model. Meanwhile, unlike traditional fixed antennas, fluid antenna systems (FAS) provide reconfigurable, position-adaptive elements that leverage spatial diversity, mitigate interference, and optimize communication performance under rapidly changing channel conditions. Despite these advantages, FL-FAS remains vulnerable to information leakage through shared model parameters. To overcome latency bottlenecks in HE-enabled FL-FAS, we formulate an uplink communication latency-aware resource allocation problem across the base station and user devices. To address this challenge, we propose FASHE-LPPO, a deep reinforcement learning (DRL) framework built upon Proximal Policy Optimization (PPO) for efficient and adaptive decision-making. The proposed approach dynamically adapts to wireless environments that vary in time while safeguarding the integrity of the FL process. Extensive simulations show that FASHE-LPPO achieves significant latency reduction and superior performance compared to benchmark optimization methods, all without sacrificing system reliability or learning accuracy.

키워드

Deep reinforcement learning; federated learning; fluid antenna system; homomorphic encryption
제목
Toward Low-Latency Privacy-Preserving Federated Learning in Fluid Antenna Systems with Homomorphic Encryption
저자
Ho, Manh Cuong; Dinh, Ton That Tam; Tran, Kiet Nguyen Tuan; Mac, Huy Dang; Oh, Junsuk; Lee, Donghyun; Song, Chihyun; Cho, Sungrae
DOI
10.1109/ICUFN69619.2026.11628633
발행일
2026-07
유형
Conference Paper
저널명
International Conference on Ubiquitous and Future Networks, ICUFN
페이지
211 ~ 216