AI-Driven Cyber Threat Prediction System Using Dynamic Graph-Based Anomaly Detection in Enterprise Networks

Enterprise networks are increasingly exposed to sophisticated cyber threats that evolve dynamically and are capable of bypassing traditional intrusion detection systems. Recent advances in Artificial Intelligence (AI) and Machine Learning (ML) have introduced new ways of analyzing network behavior using anomaly detection models. However, a major limitation in many existing approaches is the reliance on static data representations and supervised learning methods that require labeled datasets. This restricts detection of previously unseen or zero-day attacks and hinders scalability in large-scale network environments. This research proposes a novel threat prediction framework based on Dynamic Graph-based Anomaly Detection that models network hosts, connections, and communication patterns as a continuously evolving graph structure. The system employs temporal graph learning, unsupervised anomaly scoring, and behavior modeling techniques to detect abnormal communication sequences and predict potential threats. The proposed model enables real-time monitoring without requiring prior attack signatures and is capable of adapting to changes in enterprise network activity. Experimental evaluation demonstrates improved accuracy, reduced false-positive rates, and greater robustness against unknown cyber threats compared to conventional intrusion analysis. This research highlights the potential of dynamic graph learning for next-generation cybersecurity systems in large enterprise environments.