Year: 2026
Type: article
Abstract: A zero-trust deployment spanning multiple clouds may contain independently maintained authorization rules that express incompatible decisions for overlapping requests. This paper presents a compact, graph-indexed approach for detecting such cross-control-plane inconsistencies. Policies are normalized into finite sets of principals, actions, resource groups, network zones, and assurance levels, together with a cloud identifier and an allow/deny effect. An incidence graph connects policies with their canonical tokens. Candidate pairs are obtained by intersecting opposite-effect identity and resource postings, after which the remaining policy dimensions are checked exactly. The procedure reports the same conflict edges as exhaustive pairwise evaluation under the stated model. We implemented both algorithms in Python and evaluated 500–4,000 synthetically generated policies for three abstract cloud providers, using five seeds per size. In all 20 trials, the conflict sets matched. At 4,000 policies, median execution time decreased from 922.14 ms to 140.58 ms, a 6.81× median trial-level speedup, while the indexed method removed 98.43% of eligible pairs from exact verification. The results demonstrate practical candidate reduction for a transparent, bounded cross-cloud policy model.
DOI: 10.54097/6nkh8a43
Authors: Noor Bakker, Bas van der Meer
Language: English
Topic: Access Control and Trust
Subfield: Computer Networks and Communications
Field: Computer Science
Domain: Physical Sciences
Open Access status: hybrid
OpenAlex ID: W7236122279
