Sigi Watermarking Engine
Sigi is the forensic watermarking engine inside Diker. It embeds an invisible signal directly in pixel data -- imperceptible to the human eye, but machine-readable after compression, cropping, screenshots, and social media re-encoding.
How it works
Sigi modifies pixel values at a level below human perception. The embedded signal carries an identifier that links the image back to its provenance record in Diker's registry -- who created it, who licensed it, under what terms.
When metadata is stripped and the content fingerprint can't match a heavily modified derivative, the watermark is the last line of recovery.
Robustness
The watermark persists through transformations that destroy metadata and degrade content fingerprints.
Specification
| Type | Invisible image watermark |
| Supported media | Still images (JPEG, PNG, TIFF, WebP) |
| Embedding method | Pixel-domain signal injection, imperceptible to the human eye |
| Detection | Authenticated API -- submit an image, receive the provenance record if a watermark is present |
| Standards | Compatible with C2PA Content Credentials (soft binding) |
| Developer | Okeanos B.V., Amsterdam, Netherlands |
| Contact | [email protected] |
Part of the stack
Diker combines Content Credentials, content fingerprinting, and forensic watermarking. Each layer covers the gaps of the others.
Content Credentials
Cryptographic manifest embedded in the file. Tamper-evident. Stripped by platforms.
Content Fingerprint
Computed from pixels. Survives format conversion and compression. Fails on heavy edits.
Forensic Watermark
Embedded in pixels. Survives screenshots, cropping, social media. The last line of recovery.
Sigi ships as part of Diker's provenance pipeline. Talk to us about integration.