What Is Image Hashing?

Image hashing technologies range from cryptographic hashes (MD5, SHA-256) that detect exact file copies, to perceptual hashing algorithms (pHash, PhotoDNA) that detect visually similar images even after minor modifications. PhotoDNA, developed by Microsoft, is widely used by platforms to detect known CSAM and, increasingly, NCII. NCMEC's Take It Down program uses hash-sharing to help platforms automatically block re-uploads of minors' NCII. The limitation of all hashing approaches is that re-encoding an image significantly enough changes its hash, allowing perpetrators to defeat detection by minor modifications. Biometric face matching is not affected by re-encoding.

Key facts about this term

1. PhotoDNA is the standard for CSAM and NCII hash detection PhotoDNA creates a robust hash that is resistant to minor image modifications but can be defeated by more significant re-encoding. Many major platforms have implemented PhotoDNA for NCII detection.
2. Hash databases require a known original Image hashing can only detect re-uploads of known content. It cannot detect new NCII of a victim that has not been previously identified and hashed — biometric scanning fills this gap.
3. Combining hashing and biometric scanning provides the best coverage Hash-based detection prevents re-uploads of known content; biometric face scanning finds new and modified content. ScanErase uses biometric scanning as the primary detection method.