Enhancing Cyber Resilience in Cloud Environments through Graph Neural Networks and Continuous Attack Path Analysis

Abstract

Cloud computing environments are increasingly exposed to sophisticated cyber threats due to their distributed architectures, dynamic resource allocation, complex access permissions, and interconnected services. Traditional security approaches often rely on isolated vulnerability assessments and static monitoring techniques, limiting their ability to identify evolving attack paths across cloud infrastructures. This study explores the enhancement of cyber resilience through the integration of Graph Neural Networks (GNNs) and continuous attack path analysis. The proposed approach represents cloud resources, identities, applications, network connections, and security policies as a dynamic graph, enabling the identification of relationships that may facilitate unauthorized access or privilege escalation. GNNs are employed to learn structural patterns, classify potentially vulnerable entities, and prioritize high-risk attack paths, while continuous analysis updates security assessments as infrastructure configurations and threat conditions change. The methodology combines graph construction, feature engineering, supervised or semi-supervised learning, attack path enumeration, and risk-based response prioritization. Evaluation is designed around detection precision, recall, F1-score, attack path identification accuracy, computational efficiency, and resilience-related recovery indicators. The proposed framework aims to improve threat visibility, accelerate risk mitigation, and support proactive security decision-making in complex cloud environments. By integrating contextual graph learning with continuous security assessment, the study establishes a methodological foundation for adaptive, scalable, and intelligence-driven cloud cyber resilience