Secure and Efficient Intrusion Detection in MANETs: An Optimized Sparse Spectra Graph Convolutional Network With Blockchain

ABSTRACT Because they lack infrastructure and are decentralized, mobile ad hoc networks (MANETs) are intrinsically susceptible to a range of security threats because of their vital topology given the absence of centralized authority. Malicious activities in such environments can degrade network performance, deplete resources, and disrupt communication. Although several intrusion detection systems (IDS) have been proposed, most exhibit critical limitations: high detection latency, elevated computational overhead, inconsistent accuracy in evolving topologies, centralized trust mechanisms prone to single‐point failures, and insufficient modeling of complex, dynamic node interactions. In order to address these issues, this study suggests a fresh approach: secure IDS for MANETs, integrating an optimized sparse spectra graph convolutional neural network with secretary bird optimization (SSGCNN‐SBOA). The framework leverages the ISCX‐UNB dataset for training and validation. A switching hierarchical Gaussian filter (SHGF) denoises and normalizes traffic data, while the snow avalanche algorithm (SAA) enhances feature selection efficiency. To address trust issues, a blockchain framework utilizing fair proof‐of‐reputation (FPoR) consensus ensures decentralized, tamper‐resistant reputation management. The SSGCNN effectively captures spatial–temporal traffic patterns, and SBOA fine‐tunes model hyperparameters to boost the precision of recognition and reduce false alarms. Results from experiments show that SSGCNN‐SBOA is a powerful and scalable solution for secure MANET operations since it greatly exceeds current methods in terms of latency, throughput, packet delivery ratio (PDR), cost of transportation, accuracy, precision, recall, and F1‐score.