Chúc mừng Sinh viên nhóm nghiên cứu UiTiOt đã có bài báo được chấp nhận tại Hội nghị quốc tế ComNetSat 2025 được tổ chức vào 11-13/12/2025 tại Indonesia.
Tên bài báo: Performance and Security Evaluation of Federated Learning under IID and Non-IID Settings and Differential Privacy
– Sinh viên thực hiện:
- Phạm Lê Quốc Việt – ATTT2023.2 – Tác giả chính
- Đinh Công Thảo – MMTT2023.3 – Đồng tác giả
- Phạm Hoàng Anh Tú – ATTT2023.2 – Đồng tác giả
– Giảng viên hướng dẫn: PGS. TS. Lê Trung Quân và ThS. Nguyễn Khánh Thuật
Tóm tắt: Federated Learning (FL) has emerged as a promising approach for privacy-preserving machine learning by enabling multiple clients to train a global model collaboratively without sharing their raw data. Despite its decentralized nature, FL remains vulnerable to inference attacks due to the exposure of model updates. In this article, we comprehensively evaluate federated learning in IID and non-IID settings and consider the exchange between model performance and data distribution. Besides, we proposed a privacy-enhanced FL framework that integrates Gaussian-mechanism-based Differential Privacy (DP) at the client side, by injecting calibrated noise into local model parameters before aggregation. Our method ensures a tunable trade-off between utility and privacy while preserving the original FL architecture. Experiments on the CIFAR-10 dataset using the Vision Transformer (ViT-Tiny) show that while Centralized Learning (CL) achieves the highest accuracy, FL with DP can effectively preserve privacy with moderate performance overhead. These findings offer practical guidance for deploying federated learning systems that balance performance and privacy in real-world scenarios.
Homepage Hội nghị: https://comnetsat.org/
Thông tin Hội nghị: The forthcoming IEEE International Conference on Communications, Networks, and Satellite (COMNETSAT) 2025 will be held in Padang, Indonesia, on December 11–13, with a hybrid system (online and offline). IEEE COMNETSAT is a prestigious venue for academics and professionals from all over the world to exchange ideas and present the outcomes of ongoing research in the most cutting-edge areas of communications, network and information technology, satellite, broadband and photonics, data science, and artificial intelligence. We hope that the 14th IEEE COMNETSAT may build on the success of the previous IEEE COMNETSAT conferences and develop into a notable and prevalent conference year after year with the help of the IEEE Communications Society.
