Secretary Bird Red Panda Optimization Enabled EfficientNet for Intrusion Detection With Multipath Routing in Mobile Ad Hoc Networks

ABSTRACT Mobile ad hoc networks (MANETs) are decentralized wireless networks where nodes communicate without a fixed infrastructure. While MANETs are well‐suited for dynamic environments, their flexibility also exposes them to severe security threats. The principal objective of this research is to present an advanced intrusion detection model for MANETs using a novel approach termed Secretary Bird Red Panda Optimization enabled EfficientNet (SBRPO_EfficientNet). The core novelty lies in the design of the Secretary Bird Red Panda Optimization (SBRPO) algorithm, a new metaheuristic that combines the strategic attack behavior of Secretary Bird Optimization (SBOA) with the adaptability of Red Panda Optimization (RPO) to enhance multipath routing. Furthermore, EfficientNet is optimized using SBRPO for high‐accuracy intrusion detection at the Base Station (BS). The process begins with simulating the MANET and evaluating its energy and trust models. Next, multipath routing is done through the SBRPO algorithm by considering the fitness parameters including energy, delay, distance, trust, and throughput. Intrusion detection is then performed at the BS, where the input log file is normalized with quantile normalization. Then, feature selection is carried out using Bhattacharyya Distance and Wave‐Hedges metrics. The final step involves the detection of intrusion using EfficientNet, which is trained using SBRPO. The experimental results indicate that the SBRPO_EfficientNet achieved a lower delay of 0.604mS, improved energy efficiency by 3.308 J, and higher trust levels at 87.727. Additionally, the scheme demonstrated superior performance in intrusion detection, with a higher True Positive Rate (TPR), True Negative Rate (TNR), and Accuracy values of 90.913%, 90.009%, and 90.884%.