Ensemble Learning-Based Malicious Node Identification in MANETs: A Robust Approach for Attack Detection and Mitigation

MANETs are susceptible to very high rates of security attacks due to their decentralized and dynamic nature. The most challenging is detecting anti-jamming and rogue nodes, which compromise network quality and communication reliability. This research proposes an ensemble learning framework for detecting malicious nodes in MANETs. The target approach is an ensemble of heterogeneous learning methods that rely heavily on the discovery of malicious actions and observations over extended periods. The ensemble system, to achieve best possible malicious node detection accuracy and reliability, leverages diversity and complementarily among individual learning methods. Streaming communication anomaly, node activity, and network traffic activity are fed into the ensemble system. These methods, such as bagging, boosting, and stacking, are used in ensemble learning to combine the decisions of multiple models and produce a final prediction. Experimental results on real-world MANET data sets on detection accuracy and against false negatives. False positives demonstrate that the proposed ensemble learning method is better than individual learning algorithms. Experiments validate the effectiveness and practicability of an ensemble learning-based method for detecting and preventing malicious nodes in MANETs, thereby improving the network's dependability and security.