Incremental Evidence-Dependent Authorization for Software Delivery Pipelines: Design, Implementation, and Scaling Evaluation

This record contains the public preprint of the study "Incremental Evidence-Dependent Authorization for Software Delivery Pipelines: Design, Implementation, and Scaling Evaluation." The study presents an incremental evidence-dependent authorization mechanism for software-delivery pipelines that separates evidence support, capability eligibility, and held permissions; reevaluates only the affected dependency region after evidence loss; and requires an explicit restoration transition before recovered eligibility can become held authority again. The executable reproducibility artifact (Python reference implementation, synthetic scaling and conformance benchmark, runnable CI/CD-style pipeline trace, machine-readable results, deterministic logs, generated figures, and a SHA-256 manifest) is archived separately at https://doi.org/10.5281/zenodo.22882539. Under the declared positive witness-family rule class and synthetic workloads, the reported reference evaluation contains 3,300 incremental-versus-full oracle comparisons with zero state or permission mismatches and 10,000 restoration episodes with zero automatic or premature regrants. Timing measurements are environment-specific and are not claimed as universal performance results. This preprint is made public to provide a stable citation for the mechanism and its existing evaluation. It is the same research work currently under journal review; the public record does not imply journal acceptance or peer-review endorsement. The research was conducted independently by the author and does not represent official sponsorship, supervision, or endorsement by Georgetown University.