24 nov I understand that i’ve implemented it correctly because some other services who have the rule managed to make use of my hashes to precisely complement images.
Perhaps there was reasons that they do not want actually technical individuals examining PhotoDNA. Microsoft claims your "PhotoDNA hash is not reversible". That is not genuine. PhotoDNA hashes are projected into a 26x26 grayscale image which just a little blurry. 26x26 try bigger than many desktop icons; it really is enough information to recognize anyone and items. Treating a PhotoDNA hash isn't any more complicated than fixing a 26x26 Sudoku problem; a job well-suited for computers.
I have a whitepaper about PhotoDNA that I've independently distributed to NCMEC, ICMEC (NCMEC's intercontinental equivalent), a couple of ICACs, multiple tech suppliers, and Microsoft. The few just who given suggestions are extremely concerned about PhotoDNA's restrictions the paper phone calls around.