Show HN: Cerno – CAPTCHA that targets LLM reasoning, not human biology
6 hours ago
- #Proof of Work
- #Behavioral Analysis
- #Client Verification
- Proof of work using SHA-256 hash prefix with adaptive difficulty based on client signals.
- Maze generation via Growing Tree algorithm with seeded PRNG for trustless validation.
- Motor-control analysis extracting 12 behavioral features from raw pointer events, scored against per-maze baselines.
- Stroop probes implementing color-word interference at maze decision points, with timing derived from event stream.
- Signature binding using ECDSA P-256 ephemeral keypair for challenge issuance and submission verification.
- Reputation system based on behavioral consistency across sessions with EMA trust scores keyed by device identifier.
- Integration example using Cerno React component for frontend protected forms and server-side token verification.