Privacy‑Enhanced Federated Learning Framework for Intrusion Detection in Smart IoT Environments
Revolutionary Advances in Computing and Electronics: An International Journal · 2025
DOI: 10.65890/race.v1i2.153 ↗Kurz gefasst
Fachartikel in der Zeitschrift Revolutionary Advances in Computing and Electronics: An International Journal, 2025. Verfasst von Himanshu Sharma und 2 weiteren Autor:innen. Die Arbeit nennt 19 Literaturangaben. Der Volltext ist frei zugänglich (Open Access, CC BY 4.0).
Abstract (Originalsprache)
The rapid proliferation of the Internet of Things (IoT) has given rise to Smart Environments where sensors, actuators and embedded systems interact seamlessly to provide automation and convenience. Recent reports estimate that the number of connected IoT devices reached 14.4 billion in 2022 and continues to grow despite supply‑chain disruptions. Unfortunately, this connectivity also exposes critical infrastructures to malware, botnets and other cyber‑attacks. Conventional intrusion detection systems (IDS) are often ill‑suited for resource‑constrained IoT nodes because they require centralised data collection, violating privacy regulations and incurring excessive bandwidth consumption. Federated learning (FL) has emerged as a promising paradigm to overcome these limitations by enabling collaborative model training directly on edge devices. However, FL alone does not guarantee privacy, model updates may leak sensitive information and naive aggregators remain vulnerable to single points of failure. This work proposes a privacy‑enhanced federated learning framework for anomaly and malware detection in Smart IoT environments. The contributions of this research paper is threefold: (1) a hierarchical FL architecture that distributes computation across edge, fog and cloud tiers, incorporating differentially private noise to model updates is designed; (2) a multi‑agent intrusion detection algorithm that trains lightweight deep models locally using real traffic data while a fog‑level coordinator performs secure aggregation is developed; and (3) extensive experiments on modern IoT intrusion datasets to evaluate detection accuracy, communication overhead and resource consumption is done. The results show that the proposed framework achieves comparable accuracy to centralised training while substantially improving privacy and resilience.
Zitieren
APA
Sharma, H., Kumar, A., & Kumar, G. (2025). Privacy‑Enhanced Federated Learning Framework for Intrusion Detection in Smart IoT Environments. Revolutionary Advances in Computing and Electronics: An International Journal. https://doi.org/10.65890/race.v1i2.153
Vancouver
Sharma H, Kumar A, Kumar G. Privacy‑Enhanced Federated Learning Framework for Intrusion Detection in Smart IoT Environments. Revolutionary Advances in Computing and Electronics: An International Journal. 2025. doi:10.65890/race.v1i2.153
BibTeX
@article{Sharma2025,
author = {Sharma, Himanshu and Kumar, Ashok and Kumar, Gautam},
title = {Privacy‑Enhanced Federated Learning Framework for Intrusion Detection in Smart IoT Environments},
journal = {Revolutionary Advances in Computing and Electronics: An International Journal},
year = {2025},
doi = {10.65890/race.v1i2.153}
}
EUVIMED