SecureSign

Digital signature platforms are built to prove a document was signed, not to verify that the signature itself is genuine — a forged or lightly altered signature can pass through the standard e-signing flow just as easily as a real one, and nobody notices until the document is disputed later.

We treated signature and document verification as a step that happens before countersigning, not an audit that happens after a dispute. That meant building a computer-vision layer into the signing workflow itself, so a questionable signature gets flagged in the moment a counterparty is about to rely on it.

A computer-vision model trained on genuine and forged signature pairs compares each incoming signature against the signer's known reference samples, scoring it for forgery indicators — inconsistent stroke pressure, unnatural pen lifts, tracing artifacts — and a companion tamper-detection pass checks the document itself for signs of post-signature alteration, like inconsistent metadata or pixel-level edits around the signature block.

Suspicious signatures and tampered documents get flagged for review before a counterparty countersigns and relies on them, catching forgery attempts at the one point in the workflow where catching them actually prevents the dispute, rather than just providing evidence for one after the fact.