Research Article
Lightweight Intrusion Detection for IoT Systems Using Artificial Neural Networks
@INPROCEEDINGS{10.1007/978-3-031-64954-7_3, author={Radhwan A. A. Saleh and Louai Al-Awami and Mustafa Ghaleb and Anas A. Abudaqa}, title={Lightweight Intrusion Detection for IoT Systems Using Artificial Neural Networks}, proceedings={Security and Privacy in Communication Networks. 19th EAI International Conference, SecureComm 2023, Hong Kong, China, October 19-21, 2023, Proceedings, Part II}, proceedings_a={SECURECOMM PART 2}, year={2024}, month={10}, keywords={Intrusion Detection System Lightweight Internet of Things Machine Learning Artificial Neural Network}, doi={10.1007/978-3-031-64954-7_3} }- Radhwan A. A. Saleh
Louai Al-Awami
Mustafa Ghaleb
Anas A. Abudaqa
Year: 2024
Lightweight Intrusion Detection for IoT Systems Using Artificial Neural Networks
SECURECOMM PART 2
Springer
DOI: 10.1007/978-3-031-64954-7_3
Abstract
Internet of Things (IoT) systems, due to their vulnerability to a plethora of security attacks, suffer significant detrimental impacts on their reliability. Additionally, the inherent constraints in IoT devices necessitate the incorporation of lightweight security schemes endowed with the capacity to identify intrusions and serve as a robust line of defense. This study presents a lightweight Intrusion Detection System (IDS) that leverages the power of Artificial Neural Network (ANN) while addressing the challenges within IoT systems. The effectiveness of the proposed IDS has been validated through empirical testing by utilizing the ToN IoT Telemetry dataset. The proposed IDS exhibits superior performance compared to other Machine Learning-based IDS solutions identified in the literature, particularly in terms of binary classification where the feature set is continuous. The proposed IDS consistently exceeds expected benchmarks in multi-class classification, with metrics including recall, precision, accuracy, and F-score ranging between 91% and 100%. More importantly, its high accuracy and low time complexity make it an ideal choice for real-time applications, offering a superior alternative to existing AI-based IDS.
