Ask HN: Where to find a list of hashes for copyrighted material? | Hacker News Reader

Ask HN: Where to find a list of hashes for copyrighted material?

Due to the recent law the EU signed, any product is required to perform some sort of content filtering of what users upload (assuming the data is available to the public). As such, I need to perform some content filtering, but I can't seem to figure out where I'd go to gather a list of hashes of copyrighted material. Is there any API, a list of hashes or can it be requested to someone?

I'm just one software developer and don't represent any company.

View on HN

I just don't see how hashing is going to prevent copyright violations when motivated people can make insignificant edits to files to change their hashes, especially on text files (add a random string to the end), images (change the HSL of a single pixel), and music (add an extra space in the ID3 tag). The whole thing seems like a Sisyphean task. Any list of hashes will grow to unmanageable size almost immediately.

You'll need to rely on https://en.wikipedia.org/wiki/Locality-sensitive_hashing and fuzzy matching. I've heard Shazam works really well in most cases, although I'm not sure whether they submit the raw recording or just "fingerprints".

Still far from a trivial problem, especially for video.

Oh, interesting. I was unfamiliar with this, and thinking in terms of cryptographic hashes needing to be identical.

Sadly law makers aren't technical people, I don't think they have a clue on how to implement such system. Basically only big players can afford the R&D effort.

I think EU should build this as a free online service to which you send a query in some compressed & hashed form and the API would reply whether that piece is protected by copyright or not.

Go to thepiratebay.org, download a torrent, look at all the hashes within.

(note that there is basically no way to realistically filter content based on file hashes alone)

That's problematic though, it's the only reasonable implementation (reasonable as in: don't need to spend an insane amount of money on it) for any startup. Performing video/audio/images or even text comparison it's a business on its own. It sounds weird that it's not possible to get such a list

Reply on news.ycombinator.com