Public Methodology

Protection Benchmark

We publish benchmark trends so customers can track protection effectiveness over time. Full raw artifacts stay private to preserve security and model integrity.

PISR (Current)

0.0%

Baseline run, n=6 transformed variants

Quality Pass Rate

16.7%

Guardrail against visible degradation

Last Updated

March 7, 2026

Expected weekly cadence

How We Measure

Transform Suite

Protected images are tested against a fixed set of post-upload transforms (resize, recompress, crop, thumbnail, metadata strip).

Primary Metric

PISR tracks the percentage of transformed images that remain non-matchable by selected facial-recognition evaluators.

Quality Guardrail

We only promote policy changes if protection improves without dropping below image-quality thresholds.

Transparency Scope

  • Published: metric trends, methodology summary, benchmark versioning.
  • Not published: raw per-image outputs, full transform parameters, policy tuning logic.
  • Reason: protect user privacy and prevent adversarial reverse-engineering.