Building robust internet of things defense system using multi-objective nature-inspired framework

The rapid proliferation of IoT devices and the growing need for real-time processing have enhanced the quality of life, but also introduced significant security vulnerabilities. While various security solutions exist to counter malicious activities, many fail to adequately address evolving threats. Consequently, there is a clear need for an intelligent system capable of simultaneously adapting to dynamic cyber risks and improving defensive performance. Keeping this in mind, this study developed an imperative hybrid IDS framework called CNN-MGOA by integrating a convolutional neural network (CNN) and multi-objective grasshopper algorithm (MGOA) to help identify intruders in IoT-based networks. Additionally, to more comprehensively capture the important features of network intrusion and improve the detection performance, the multi-objective grasshopper optimization algorithm (MGOA) is utilized in conjunction with a multi-class support vector machine. In this work, comprehensive experiments are conducted on three up-to-date benchmark datasets, including ToN-IoT, CIDD, and NSL-KDD. Experimental results show that our model achieves high detection accuracy of 99.89%, 99.98%, and 99.86%, in NSL-KDD, ToN-IoT, and CIDD datasets, respectively, while obtaining low False Positive Rates of 0.012, 0.049, and 0.041. In addition, it achieved a better effect on anomaly detection than existing state-of-the-art anomaly detection systems. Despite its promising performance, our ability to detect novel risks can be constrained by the presumption of static traffic patterns and the failure to integrate behavioural characteristics. In the near future, we can employ dynamic context-aware analysis to enhance anomaly identification in dynamic IoT scenarios, thereby overcoming existing limitations.