A Comparative Analysis of Deep Learning Algorithms for Intrusion Detection in IoT
Cherukuri Gaurav Sushant, Veeramraju Lakshmi Ajay, Rashmi Sahay · 2024
The increase in the use of the Internet of Things (IoT) to achieve concepts like smart cities and health monitoring brings together a diversity of devices and data flow, which creates an abyss of a diverse range of attacks, which makes it necessary to put in place proper anomaly identification and mitigation mechanisms. Traditional intrusion detection systems cannot cope with versatile attacks, which urges using advanced techniques encompassing concepts like machine learning and deep learning to create a robust and versatile mechanism to identify and mitigate unintended traffic. In this paper, we analyze the efficiency of different deep learning techniques like CNN, LSTM, and XGBoost and compare their efficiency. The paper uses a dual CNN approach, an LSTM with three stacked layers, and XGBoost techniques with cross-validation to reduce overfitting on the IOTBOTNET 2020 dataset. The results show LSTM and XGBoost perform better, achieving an f1 score of 99.97% and 94%, respectively, compared to DUAL CNN, which achieved a score of 62.49 %.