Perceptual hashing is a technique that generates a compact digital fingerprint (hash) of an image based on its visual content, not its exact pixel data. Unlike cryptographic hashing (which changes completely with any pixel modification), perceptual hashing produces similar hash values for visually similar images. This allows platforms to detect modified images that may appear in unauthorized uses (slightly cropped, color-adjusted, or otherwise altered to evade exact-match detection). Microsoft's PhotoDNA is the most widely deployed perceptual hashing system, used by Meta, Google, and hundreds of other platforms to detect unauthorized exposure. ScanErase utilizes perceptual hashing in its discovery processes to enhance exposure detection on supported platforms.

Key facts about this term

  1. Understand that hash matching helps detect similar content Once your content's hash is in a platform's database, similar or closely resembling uploads can be detected.
  2. ScanErase uses hash matching in discovery processes Hash matching is a standard component of ScanErase's processes for platforms that support this feature.
  3. Recognize that modified images may evade simple hashing For heavily modified images, additional discovery techniques may be needed, ScanErase's scanning capabilities assist with this.

Frequently asked questions

Is perceptual hashing the same as the biometric scan ScanErase performs?

No. Perceptual hashing matches images based on visual content similarity. ScanErase's biometric scan uses facial geometry embeddings, a different and more effective technique specifically designed to find faces across AI-generated variations.

What happens if someone slightly modifies my images to evade hash detection?

Moderately modified images may evade perceptual hash detection. ScanErase's biometric scan uses facial geometry, not image hashing, so it captures most modifications that maintain facial structure.

Related glossary terms