Abstract
A statistical safety gate is described herein for automatically deploying machine-tuned configurations to a production network controller, including a traffic-pattern-correlation-based link-detection engine. The gate compares a candidate engine configuration with a current production configuration using paired statistical hypothesis testing. It automatically selects McNemar's test for normal sample sizes, an exact binomial test for small discordant-pair counts, or an explicit bypass when a sample is too small. A short-circuit prevents unnecessary configuration changes when the current configuration already satisfies an operator-defined accuracy threshold. A tuned configuration is deployed only when statistical significance and positive directional improvement are both present. Candidate configurations that are not deployed are persisted with evaluation results and decision data for operator review, later re-evaluation, and auditing. The gate therefore protects production networks from noise-driven configuration churn while allowing genuine optimization improvements to proceed automatically, and can be applied more broadly to networking systems that infer topology from incomplete or noisy data.
Creative Commons License
This work is licensed under a Creative Commons Attribution 4.0 License.
Recommended Citation
Dec, Wojciech and Menashe, Ori, "METHOD FOR IMPROVING THE STABILITY AND ACCURACY OF NETWORK CORRELATION ENGINE OPTIMIZATION", Technical Disclosure Commons, (September 16, 2026)
https://www.tdcommons.org/dpubs_series/11759