Belarbi, Othmane, Pramadi, Yogha Restu, Rana, Omer  ORCID: https://orcid.org/0000-0003-3597-2646, Li, Yuhua ORCID: https://orcid.org/0000-0003-2913-4478, Khan, Aftab and Spyridopoulos, Theodoros ORCID: https://orcid.org/0000-0001-7575-9909
2025.
FedHDC-IDS: a hyperdimensional computing based network intrusion detection dystems for IoT networks.
Presented at: 2025 IEEE 45th International Conference on Distributed Computing Systems Workshops (ICDCSW),
Glasgow, United Kingdom,
21-23 July 2025.
2025 IEEE 45th International Conference on Distributed Computing Systems Workshops (ICDCSW).
2025 IEEE 45th International Conference on Distributed Computing Systems Workshops (ICDCSW).
IEEE,
pp. 677-682.
10.1109/ICDCSW63273.2025.00126
|
Abstract
Traditional AI-based Network Intrusion Detection Systems (NIDS) typically rely on centralised analysis of collected data, which is impractical and raises privacy concerns in distributed IoT networks. Federated Learning (FL) has been explored as a method to preserve privacy, enabling decentralised model training in AI NIDS. Although FL minimises data transfers, it still suffers from high communication overhead and increased computational costs that emerge from the underlying Deep Learning (DL) model. FedHDC-IDS, a novel approach that combines Federated Learning (FL) and Hyperdimensional Computing (HDC) to enhance the security of IoT networks through efficient collaborative learning while preserving user privacy. HDC offers lightweight, energyefficient learning with fast convergence, making it well-suited for resource-constrained IoT devices. Our framework uses Radial Basis Function (RBF) encoding to efficiently represent data and FedAvg for model aggregation.
| Item Type: | Conference or Workshop Item (Paper) |
|---|---|
| Date Type: | Publication |
| Schools: | Schools > Computer Science & Informatics |
| Publisher: | IEEE |
| ISBN: | 979-8-3315-1725-0/25 |
| ISSN: | 2332-5666 |
| Last Modified: | 10 Dec 2025 09:45 |
| URI: | https://orca.cardiff.ac.uk/id/eprint/183016 |
Actions (repository staff only)
 |
Edit Item |
Altmetric
Altmetric