Deep Learning and Optimization Approaches in Similarity-Navigated Graph Neural Networks and Lightweight Cryptography for Preventing Black Hole Attacks in MANET: A Review

Abstract

Mobile Ad Hoc Networks (MANETs) are decentralized and infrastructure-less wireless networks that enable dynamic communication among mobile nodes. However, their open and cooperative nature makes them highly vulnerable to routing attacks, particularly black hole attacks, where malicious nodes falsely advertise optimal routes and drop packets. This paper presents a comprehensive review of deep learning and optimization-based approaches, focusing on Similarity-Navigated Graph Neural Networks (SNGNN) and lightweight cryptographic mechanisms for securing MANETs. Graph Neural Networks (GNNs) have emerged as a powerful tool for modeling network topology and detecting anomalous node behavior through relational learning. The integration of similarity navigation enhances node embedding by capturing structural and feature-based similarities, improving attack detection accuracy. Furthermore, lightweight cryptography ensures secure communication with minimal computational overhead, making it suitable for resource-constrained MANET environments. Recent studies (2020–2025) demonstrate that combining GNN-based detection with optimization algorithms significantly enhances detection rates (above 95%) while maintaining network efficiency. However, challenges such as energy consumption, scalability, and real-time deployment remain critical. This review analyzes current techniques, identifies research gaps, and highlights future directions for developing efficient and secure MANET systems.