I think you can divide it (among many other ways) into two categories: known content, and unknown content. PhotoDNA solves the known content (hashed) problem. T

I think you can divide it (among many other ways) into two categories: known content, and unknown content. PhotoDNA solves the known content (hashed) problem. The other stuff you mentioned I believe is being leveraged to combat the unknown [abuse] content problem. Eg, identifying the "0-day" content.