Muhammad M, Arumugam S, Sardy LA (2025)
Publication Type: Conference contribution
Publication year: 2025
Publisher: Institute of Electrical and Electronics Engineers Inc.
Pages Range: 199-206
Conference Proceedings Title: 2025 9th Cyber Security in Networking Conference: GenAI and Cybersecurity, CSNet 2025
Event location: Abu Dhabi, ARE
ISBN: 9798331575564
DOI: 10.1109/CSNet67572.2025.11288241
Intrusion Detection Systems (IDSs) play a crucial role in securing Internet of Things (IoT) networks, which are increasingly exposed to sophisticated cyber threats. This paper presents an adaptive, reconstruction-based IDS leveraging multiple class-specific Long Short-Term Memory Autoencoders (LSTM-AEs), each trained on a single traffic class (Normal, DDoS-HTTP, DDoS-TCP, DDoS-ICMP). Unlike conventional anomaly detection that models only normal traffic, our approach performs multi-class classification by comparing reconstruction errors across all class-specific models. Evaluated on the Edge-IIoTset dataset, the method achieved a macro-averaged F1-score of 0.9963, underscoring the suitability of LSTM-AE architectures for fine-grained, threat-specific detection in realistic IoT environments.
APA:
Muhammad, M., Arumugam, S., & Sardy, L.A. (2025). AI-Based Anomaly Detection for Enhanced Cybersecurity in IoT Networks. In 2025 9th Cyber Security in Networking Conference: GenAI and Cybersecurity, CSNet 2025 (pp. 199-206). Abu Dhabi, ARE: Institute of Electrical and Electronics Engineers Inc..
MLA:
Muhammad, Mamdouh, Sushmetha Arumugam, and Loui Al Sardy. "AI-Based Anomaly Detection for Enhanced Cybersecurity in IoT Networks." Proceedings of the 9th Cyber Security in Networking Conference, CSNet 2025, Abu Dhabi, ARE Institute of Electrical and Electronics Engineers Inc., 2025. 199-206.
BibTeX: Download