Hash functions are algorithms applied to inputs of variable length to provide a fixed-length output. A given hash algorithm (such as SHA-256) always returns the same value for a given input, making it a means of uniquely identifying a piece of digital content (such as an image, video, or block of text).
See also: Algorithm
Commentary:
- Content hashes therefore form the basis of hash filtering, whereby items exactly matching known hashes may be detected and acted on automatically, triggering actions such as a prohibition from being uploaded or served by a system.
- A limitation of conventional hashing is that minor modifications to inputs will result in new unique hashes, thereby allowing nearly identical copies to defeat filtering based on previously known hashes.
- Perceptual hashing is a technique of detecting content that is substantially similar to known hashes, which may be escalated to human review or automatically blocked based on predefined thresholds of similarity.