Face-scan Verification & Fraud Console

Face-scan verification, step by step

• Detects the face and turns it into a 128-number signature with OpenCV (YuNet and SFace).
• Runs a reverse image search on Google Lens and Yandex at once, and lists the social platforms the face appears on.
• Confirms the top match is the same person by comparing the two signatures (cosine similarity of 0.363 or more).
• Takes a SHA-256 fingerprint of a canonical record of that post and writes it to Solana devnet with the SPL Memo program.
• Reads it back, re-hashes and shows VERIFIED, then proves that changing a single character is caught as tampering.

Why SHA-256 and Solana

The same record always gives the same 64-character hash, and one changed character changes about half of it. A memo transaction costs a fraction of a cent, confirms in under a second, needs no smart contract, and anyone can check it on Solana Explorer. Only the hash goes on chain, never the personal data.

Fraud investigation console

An agentic console that investigates card transactions held in TigerGraph and lists them as fraud, uncertain or legitimate.

Team

Both are team builds from Hacker House Goa 2026, and the face-scan tool was made with my team, ASTRA-X. At the same event I built VANTAGE: ZERO solo for the Wispr Flow challenge.