A Comprehensive Review of Hybrid Transformer based Gated Graph Attention Capsule Network Design for Preventing Attack in Radar Target Detection

Abstract

Radar target detection systems are essential in modern defence, surveillance, and autonomous applications, but they are increasingly vulnerable to adversarial attacks, signal interference, spoofing, and jamming. These threats can significantly degrade detection accuracy and compromise system reliability. To address these challenges, artificial intelligence (AI), particularly hybrid deep learning architectures, has emerged as an effective solution for secure and robust radar signal processing. This paper presents a comprehensive review of hybrid Transformer-based gated graph attention capsule network (TGACN) architectures for preventing attacks in radar target detection systems. The integration of Transformers, Graph Attention Networks (GAT), and Capsule Networks enables efficient extraction of spatial, temporal, and relational features from complex radar data. Transformers provide global contextual understanding, GAT models capture interdependencies between targets and signals, and Capsule Networks preserve hierarchical feature representations while improving robustness against adversarial perturbations. Recent studies demonstrate that such hybrid models significantly enhance detection accuracy, adaptability, and resilience under noisy and adversarial conditions. Despite these advancements, challenges remain in terms of computational complexity, large-scale data requirements, and real-time implementation. This review highlights recent developments, key methodologies, and limitations, and outlines future directions including lightweight models, explainable AI, and edge-based deployment for improved radar system security.