Integrated intrusion detection design with discretion of leading agent using machine learning for efficient MANET system

MANET is a hot research subject. Its qualities, including no infrastructure, fast network setup, and no centralized management, have led to its popularity and widespread use in numerous sectors. A major part of the network is security. Intrusion Discovery Scheme (IDS) is a network security strategy. In this paper, we implemented effective intrusion detection and efficient clustering with cluster head selection. For these two stages of implementation, we are not integrating any logic combinations to make decisions. Machine learning models are filled up that place of work in the proposed optimal MANET design. At first, the IDS is performed in the network using Adaptive Ensemble Tree Learning (AETL) based classification of typical nodes and malicious intrusions. Once the attacker nodes are identified, the nodes will be recovered to mold the network for the next sequence process of clustering. In the disemboweled network, the proposed model of second stage Hybrid Dual Optimization of Machine Learning Model (HDOMLM) is applied to elect the leading agent node in the formed clusters. Particle Swarm Optimization (PSO) is defined for the initial clustering of nodes and immediately the O-MLM is performed to detect the leading agent nodes in each cluster with the selection features of node degree, node mobility, energy, distance and delay. Experiment validations are accomplished to analyze the results of the proposed method using the MATLAB simulation tool and quantitative evaluations done for different Key Performance Indices (KPI) of network lifetime, residual energy, packet delivery and transmission delay with earlier works of the same. From the simulation performances, our proposed AETL with HDOMLM attained the peak results than other algorithms with the metrics of 71% of energy saving and 50.12% enlarging the network lifetime.