Intrusion Detection System in IoT Smart city Environment using Tree-based Approach with Swarm-based Optimization for Multi-Step Cyber-Attack Dataset

With the popularity of IoT technologies, security has become a significant concern. IDS is a crucial component that analyses IoT network activity and categorizes it as either normal or anomalous due to susceptible attacks from hackers. An IDS safeguards a system's availability, integrity, and confidentiality. Because of resource constraints and algorithmic complexity, traditional ML IDS have drawbacks when applied to IoT networks. A lightweight IoT IDS that is optimal, inexpensive, and can minimize the loss function. So, to find the optimal model parameter values of the ML algorithm, hyperparameter tuning is performed by using populationbased swarm algorithms. The swarm-based algorithms control the learning process to extract the ML algorithms' relevant parameters and ultimately learn. A comparative study between the proposed hyperparameter tuning and default parameters predictions is made. The experimental results confirmed that the suggested strategy improves the security of the IoT environment. This approach is evaluated and determined to be promising on the Multi-Step Cyber-Attack (MSCA) benchmark IDS dataset. Additionally, the suggested method outperforms standard Tree-based ML parameters in accuracy while achieving a satisfactory trade-off between effectiveness and productivity.