Exécution de notre outil de synthèse sur un nouvel article

Vérifiez l’article ci-dessous, puis lancez sa synthèse.


Detecting False Assurance in Cloud Control Inheritance Using a Federated Compliance Integrity Graph

Arun Chowdary Sanna

IEEE ... International Conference on Cloud Computing
Licence : non indiquée (tous droits réservés)

Résumé : Cloud programs routinely rely on inherited security controls from shared platform providers, yet inherited assurance is typically maintained through static documentation and periodic manual reviews. This creates a gap between claimed compliance and actual control effectiveness when cloud-native architectures drift. We present the Federated Compliance Integrity Graph (FCIG), a time-aware directed property graph that links controls, platform components, system boundaries, evidence artifacts, and assessment outcomes across Common Control Provider (CCP) and System Owner domains. FCIG defines quantitative metrics for evidence freshness, scope-aware coverage, inheritance validity, and false assurance rate. We implement a prototype pipeline using machine-readable OSCAL artifacts and runtime evidence adapters, and evaluate it against fourteen reproducible drift scenarios in a Kubernetes-based cloud-native environment. FCIG achieves perfect recall (R = 1.0, micro-F1= 0.624) across all positive scenarios, where the strongest baseline (precision-optimized limited mapping) achieves F1= 0.51 at R = 0.38, and full-mapping baselines achieve F1≤0.15, with sub-minute detection latency (p50=0.27 min) versus 72-hour manual review cycles. The precision trade-off (P = 0.45) reflects conservative transitive dependency analysis, which is appropriate for compliance monitoring where false negatives carry greater risk than additional alerts. Three independently-derived extension scenarios test detection generalization beyond the co-designed rule set. Holdout variants with topology noise confirm robustness, and four negative controls (including two scope-aware negatives that emit signals suppressed by boundary annotations) produce zero false positives on benign or authorized changes. The work operationalizes RMF-aligned continuous monitoring by converting inherited controls from static assertions into continuously verifiable claims.

Publié le 01 Jul. 2026