With the rapid growth of cloud computing, IoT devices, and distributed networks, cyberattacks have become increasingly sophisticated and difficult to detect using conventional intrusion detection systems (IDS). Traditional machine learning approaches often struggle to capture complex temporal and structural relationships within network traffic data. This paper proposes a novel Hybrid Transformer-Graph Neural Network Intrusion Detection System (HTGNN-IDS) that combines Transformer-based temporal feature extraction with Graph Neural Network-based relational learning. Additionally, Explainable Artificial Intelligence (XAI) techniques are integrated to provide interpretable threat analysis for cybersecurity analysts. The proposed framework is evaluated on benchmark intrusion datasets including CICIDS2017, UNSW-NB15, and CSE-CIC-IDS2018. Experimental results demonstrate superior detection accuracy, precision, recall, and F1-score compared to state-of-the-art machine learning and deep learning methods. The proposed model achieves 99.12% accuracy while maintaining low false alarm rates and enhanced interpretability.

Keywords: Intrusion Detection System, Artificial Intelligence, Deep Learning, Transformer Networks, Graph Neural Networks, Explainable AI, Cybersecurity.