Optimizing IoT Security:Advanced Intrusion Detection for IoT Networks Using Deep Learning Techniques
Khandakar Rabbi Ahmed, Nusrat Ameri, Jannat Rosul Nisha, Aminul Hoque, Towfika Salam, Md Istiak Hasan Rial · 2024
The advancement of Internet of Things (IoT) devices has ushered in a new era of connectivity and convenience. Intrusion Detection Systems (IDS) are vital for protecting IoT networks by identifying abnormal activities indicative of malicious intent. Network systems face Distributed Denial of Service (DDoS), Denial of Service (DoS), and service scans, requiring sophisticated IDS that can quickly detect and neutralize these threats. Current IDS solutions need accuracy, efficiency, flexibility, and scalability improvements to detect attack traffic across various IoT networks effectively. The evolving nature of these attacks necessitates the use of up-to-date datasets. This paper contributes to IoT security by developing and assessing an IDS using the BoTNeTIoT-L01 dataset under Mirai and Gafgyt botnet attacks and different sub-type attacks. We evaluate several machine learning models: Decision Trees (DT), Random Forests (RF), LightGBM, and Multilayer Perceptron (MLP). Our results show that MLP outperforms other models, achieving a precision of $\mathbf{9 6. 4 8 \%}$, recall of $\mathbf{9 6. 2 4 \%}$, $\mathbf{F}$-score of $\mathbf{9 6. 7 2 \%}$, and accuracy of $\mathbf{9 6. 6 5 \%}$. This approach ensures our defence strategies remain effective and responsive to the latest attack vectors.