ThreatExchange

It’s somewhere in the FAQ at https://stopncii.org/faq/, but It’s PDQ for photos (https://github.com/facebook/ThreatExchange/tree/main/pdq) and MD5 for videos. PDQ is resistant to some modifications (it focuses on the ones that come from regular usage, such as changing the format from gif to jpg, or a filter changing colors or brightness), but it’s not as resistant to modifications as you could get by training dedicated classifiers or other approaches that you might do with the original media or by storing more context, which StopNCII chose not to do.

Yup, PDQ (https://github.com/facebook/ThreatExchange/tree/main/pdq) and MD5- https://stopncii.org/faq/. MD5 is a cryptographic hash, which means that even a single bit changed gives you an entirely different hash. This usually precludes any attacks to try and deliberately generate collisions, but also means it's harder to match “benign” changes - for example, many platforms re-encode videos as part of the upload process (reducing resolution, changing formats, etc). However, many platforms skip re-encoding, so MD5 can have better results than you might think at first glance. The more bits you take from the original content, the harder you can make it to bypass, but the more you might have to worry about capturing enough that you might compromise the privacy of the original submitter. For the program, StopNCII picked a set of tradeoffs betw

Hey everyone! I’m a software engineer at Meta, and I helped work on the program that the newsroom post is about. Sorry I’m late, I only just found the thread! I might be able to answer some questions about the technology - many of the building blocks are open source, like the PDQ photo hashing algorithm (https://github.com/facebook/ThreatExchange/tree/main/pdq) mentioned in FAQ (https://stopncii.org/faq/). I might not be able to answer every question, but I’ll do my best!

The FAQ (https://stopncii.org/faq/) mentions its using PDQ (https://github.com/facebook/ThreatExchange/tree/main/pdq) for images, and MD5 for videos. Companies would then use hashes to scan their platform. Meta also has a post on it: https://about.fb.com/news/2021/12/strengthening-efforts-agai...

Notably, none of the the algorithms tested in the cited study are Apple's NeuralHash, or comparable algorithms. They look at aHash, pHash (plus a variant thereof), dHash, and and PDQ (used at Facebook for similar applications, apparently). The first 4 date to between 2004 and 2010; the last is more recent, but conceptually similar - the citation [0] for PDQ puts it in the same bucket of 'shallower, stricter, cheaper, faster' algorithms as the first four. No one has proposed any of those as 'illegal image detectors'. Apple's NeuralHash may or may not be robust to the same or different perturbations but the cited study provides basically no new information to inform the conversation its press release wants to be a part of. [0]: https://github.com/facebook/ThreatExchange/blob/main/hashing...

The non-VentureBeat press release: https://newsroom.fb.com/news/2019/08/open-source-photo-video... and the relevant GitHub link: https://github.com/facebook/ThreatExchange/tree/master/hashi...