GRAYPASSDOC/ARCH-01
The pipeline,
run on you.
GrayPass verifies identity from how users interact, not from what they remember. The scene below carries your own cursor through the four stages: raw signal, feature vector, salted print, decision.
I. THE PIPELINE
01 · Raw signal
It starts with how you move.
Reaction timing, keystroke intervals, pointer dynamics. The runtime reads rhythm, never content. The trace here is your own cursor, live.
v 0.00 px/ms · jitter 0.00 · curv 0.00
II. AROUND THE PIPELINE
The rest of the system, in plain language.
Security and privacy
Everything at rest is encrypted, and everything kept is a salted print, never behavior. Multiple independent controls sit between the front door and a decision, and any research data is opt-in, pseudonymized, and revocable.
Telemetry
Session events carry confidence and reason codes for reliability and abuse defense. Aggregated counters track drift, throughput, and decision health. Operators can request curated evaluation views by browser, cohort, or mode without exposing private inputs.
DECISION STREAM · REASON-CODED
accept · conf 0.97 · print.match
accept · conf 0.94 · cadence.stable
deny · replay.nonce_mismatch · conf 0.03
accept · conf 0.96 · rhythm.consistent
review · conf 0.71 · drift.recovering
EVERY ROW AUDITABLE · NO SILENT DECISIONS
Model operations
Models are tuned per user, and every score is calibrated so policy reads in human numbers: accept above this confidence, review below it. Updates ship without downtime. The full methodology is available to security teams under review.
SEED ROTATES · TRAFFIC NEVER DROPS · 0 DOWNTIME
Why it works
A privacy-preserving print plus calibrated similarity gives a system that is fast on easy cases, tolerant of human variance, and explicit about uncertainty. That is the bar we ship against.
Want this wired into your stack?
We set time aside for engineering teams.