AI-Driven Quantitative Cybersecurity Risk Assessment of Identity and Access Management in Hybrid Cloud Environments
DOI:
https://doi.org/10.63125/1yc7ee20Keywords:
Identity Management, Hybrid Cloud, Cybersecurity Risk, Artificial Intelligence, Access ControlAbstract
Identity and access management (IAM) represents a central cybersecurity control in hybrid cloud environments, where identities, credentials, and permissions operate across on-premises, private-cloud, and public-cloud systems. This study was designed to quantitatively assess IAM-related risk factors and evaluate the performance of artificial intelligence (AI) in identifying risk-related events. A retrospective longitudinal observational approach was specified using identity, authentication, authorization, and security records collected over a defined observation period. The dataset included 1,200 identities, comprising 900 human and 300 non-human identities, and 3.58 million analyzed authentication and authorization events across three environment types. It included 240 confirmed compromise records. Multivariable logistic regression was used to examine associations between IAM characteristics and compromise outcomes. AI model performance was compared with a rule-based baseline on a held-out sample of 300 identities, including 60 with confirmed compromise. In the findings, compromise was recorded for 20.0% of identities. Lack of multi-factor authentication was associated with higher odds of compromise (adjusted odds ratio [OR] = 2.14, 95% confidence interval [CI] [1.53, 2.99], p < .001), as was privileged access (adjusted OR = 2.37, 95% CI [1.67, 3.36], p < .001). Public-cloud identity status did not differ significantly from on-premises status (adjusted OR = 1.08, 95% CI [0.73, 1.59], p = .70). The AI model achieved a receiver operating characteristic area under the curve of .91, compared with .82 for the rule-based baseline. Its sensitivity was 80.0%, while the baseline sensitivity was 60.0%; the baseline showed higher precision, at 94.7% compared with 90.6%. These example results illustrate how quantitative IAM indicators and AI performance could be reported together. They do not establish causal effects or represent empirical findings. Actual conclusions require verified dataset values, model validation, and assessment of data completeness across environments.


