Abstract
Cloud computing provides on-demand services and comprehensive security issues of multi-tenant, complex, and dynamic services. Existing threat detection techniques do not perform well in real-time and predictive risk management. This paper describes a machine learning-based solution to detect and assess threats and risks in the cloud environment in a real-time. The architecture combines four new algorithms: Temporal Anomaly Correlation Network (TACN) for identifying sequential threats, Adaptive Contextual Risk Evaluator (ACRE) for dynamically scoring risks, Hierarchical Behaviour Fusion Algorithm (HBFA) for analysing multi-layer behaviour, and Predictive Threat Propagation Simulator (PTPS) for predicting threat propagation. The framework is evaluated on a large-scale cloud dataset, demonstrating high accuracy, low false positives, and actionable risk insights. The suggested solution will improve proactive cloud security, aid intent-based decision-making, and facilitate effective countermeasures against the threat within multi-faceted cloud infrastructures.