ABSTRACT
Strong security methods are now essential given the explosive growth of the Internet of Things (IoT) and the rise in popularity of mobile ad-hoc networks (MANET). MANETs must be protected from a variety of security threats, and intrusion detection systems (IDS) are essential for this. MANETs are mobile node networks that can set themselves up and connect with one another without a centralized infrastructure. However, due to their decentralized structure and changeable topology, they are open to a number of security threats. IDS systems are crucial for real-time detection and mitigation of these assaults. Most of the literature on detecting intrusions bases its conclusions on the route request (RREQ) creation attribute to a particular node. These algorithms function well up to a point, but they have a high percentage of false positives, which reduces the network's effectiveness. A significant research challenge is creating an effective IDS system from a heterogeneous dataset. This study pre-processes the CSE-CIC-IDS2018 dataset, selects features using Cat Swarm Optimization, and then applies the XGBoost approach to classify MANET intrusion. After classification, the best choice for XGBoost's weight is made via grasshopper optimization (GOA). When compared to all other models, the proposed model performed better, scoring 98% accuracy.
