A multi-objective secure multicast routing architecture using multi-scale dilated LSTM and hybrid coati–dolphin optimization in MANET environments

In this research work, an innovative MD-LSTM (Malicious Detection Long Short-Term Memory) framework has been proposed for the accurate detection of malicious nodes and a Hybrid Coati-Dolphin Swarm Optimization (HCDSO) algorithm for secure multicast routing in MANETs. The research aims to address major MANET challenges, including dynamic topology changes, routing vulnerabilities, and malicious attacks that affect network security and communication reliability. The MD-LSTM model uses a dilation-aware recurrent memory mechanism and multi-scale convolutional feature extraction to capture long-range temporal dependencies and complex attack patterns from network traffic data, thus improving the accuracy of malicious activity detection. The HCDSO algorithm combines the global exploration capability of the Coati Optimization Algorithm (COA) with the local exploitation capability of the Dolphin Swarm Optimization (DSO) to determine optimal routing paths that maximise throughput and PDR while minimising delay, hop count, and energy consumption. The experiments were conducted on three standard datasets, namely CICIDS-2017, NSL-KDD, and KDD Cup 99, and compared against conventional deep learning models (NN, DNN, SVM, and LSTM) as well as the latest state-of-the-art models. MD-LSTM achieved maximum accuracies of 97.802%, 96.905%, and 97.807% across the three datasets, respectively, representing performance improvements of 10.12–18.12% over conventional classifiers. The HCDSO routing framework increased throughput by 8.14%, 9.13%, 10.78% and 14.28% compared with DSO, COA, POA and BCO, respectively. The findings indicate that the proposed integrated framework can significantly improve malicious-node detection accuracy, routing efficiency, network reliability, and security, providing a strong foundation for future MANET applications.