A hybrid approach for detecting network layer attacks in MANET

Mobile Ad Hoc Networks (MANETs) are crucial in dynamic and infrastructure-less environments such as disaster recovery, military operations, and intelligent transportation systems. However, their decentralized nature, high mobility, and limited computational resources make them highly susceptible to security threats, particularly black hole and wormhole attacks. This study introduces an optimized security framework that integrates the Improved Gannet Optimization Algorithm (IGOA) with a Deep Recurrent Neural Network (DRNN) for effective detection of malicious nodes in MANETs. The framework employs the Ad hoc On-Demand Distance Vector (AODV) routing protocol to establish optimal communication paths, while historical transaction data are analyzed to identify attacker nodes. The DRNN is trained on these extracted features to classify normal and malicious nodes, and its weighting factors are optimally tuned using IGOA to enhance detection accuracy. The proposed methodology is implemented on the NS2 simulation platform and evaluated against conventional techniques such as Deep Neural Networks (DNN) and DNN-Particle Swarm Optimization (DNN-PSO). Performance metrics, including accuracy, precision, recall, and F-measure, validate the effectiveness of the proposed approach, demonstrating improved attack detection rates while minimizing false alarms. The results highlight the robustness of the model in mitigating MANET security threats, ensuring reliable and secure communication. Limitations of the proposed approach include potential scalability challenges and real-time processing overhead in highly dynamic MANET environments. Despite these constraints, the model demonstrates superior attack detection accuracy, making it a viable solution for enhancing MANET security. The proposed approach outperforms DNN-PSO and DNN by achieving higher throughput (52,014 kbps for 80 users) and lower delay (11.141 s for 4 attacks).