Researchers produce collision in Apple’s child-abuse hashing system
Illustration by Alex Castro / The Verge Researchers have produced a collision in iOS's built-in hash function, raising new concerns about the integrity of Apple's CSAM-scanning system. The flaw affects the hashing system, called NeuralHash, which allows Apple to check for exact matches of known child-abuse imagery without possessing any of the images or gleaning any information about non-matching pictures.
On Tuesday, a GitHub user called Asuhariet Ygvar posted code for a reconstructed Python version of NeuralHash, which he claimed to have reverse-engineered from previous versions of iOS. The GitHub post also includes instructions on how to extract the NeuralMatch files from a current macOS or iOS build.
Early tests show that it can tolerate image resizing and compression,...