A Secure Routing and Intrusion Detection System Using Genetic Siberian Tiger Optimization ‐ Based Hybrid Deep Learning Approach in MANET

Mobile Ad Hoc Networks (MANETs) are self‐organizing wireless networks categorized by dynamic topology, high susceptibility to malicious attacks, and decentralized control. Existing studies have mainly addressed secure routing and intrusion detection distinctly, resulting in inadequate adaptability in highly dynamic and resource‐constrained environments. To address these limitations, this research proposes a unified secure routing and intrusion detection framework based on Genetic Siberian Tiger Optimization (GSTO) combined with a hybrid deep learning model. The GSTO algorithm accomplishes optimal route selection by considering multiple quality‐of‐service parameters like node energy, distance, link quality, and trust, thereby confirming secure and energy‐efficient communication. For intrusion detection, non‐negative matrix factorization (NMF) is used for dimensionality reduction and feature refinement, pursued by a depthwise separable convolutional gated recurrent unit (DWC‐GRU) model to efficiently obtain spatial and temporal dependencies in network traffic with condensed computational cost. The incorporation of optimization, feature selection, and lightweight deep learning improves scalability, diminishes false alarm rates, and advances routing consistency. Experimental outcomes on the UNSW‐NB15 and NSL‐KDD datasets validate that the proposed model attains detection accuracies of 99.55% and 99.64%, correspondingly. Along with condensed false positive rates and lower energy consumption, the proposed model provides a scalable and secure solution for MANET environments.