> *it's trivial even for non-techies to circumvent hash-based detection

I was under the impression that PhotoDNA was more than simple hashing - that it used an algorithm like Google's "similar images"