Comment by djsumdog | Hacker News Reader

Comment by djsumdog | Hacker News Reader

Parent Full thread

Yea, they filter their e-mail platforms for sure, but part of it might be the justice department binding their hands.

People in companies that track this stuff down get special hash lists from certain law enforcement agencies and the evidence itself is pretty much a controlled substance; so they need special authorization to safely handle and report it with a chain of custody. Two Microsoft employees got PTSD because of the large amount of data they had to check:

https://www.telegraph.co.uk/news/2017/01/12/two-former-micro...

I knew a lawyer who had to defend someone in this type of case. He was allowed to view the evidence (he chose not to), but only him. No paralegals, no one else in his office that wasn't directly sitting on the defense table.

If you're not granted specific authorization, even looking at this content is illegal, so Microsoft, Facebook, Apple, Google, etc have a very limited staff of people who are even authorized to handle this stuff.

It's a difficult problem.

They don't just filter their e-mail platforms and consume hash lists from law enforcement agencies. On the contrary, Microsoft funds and freely distributes the most common tool for checking photos (not file hash based, I think proprietary). That's likely the project the employees you mentioned were on, a job I wouldn't wish on anyone I know. I hope they were well supported both while working and afterward.

https://www.microsoft.com/en-us/photodna

It states immediately below the fold that it uses hash lists.

I want to avoid using the term - even though they use it - because it's not a hash function like most people are familiar from computer science. Their "hash" function is resistant to alterations and is more like a "fingerprint".

Edit: to be completely clear, PhotoDNA isn't a cryptographic hash. It's a hash function that maximizes similarity of hashes based on inputs, and is probably closer to a bloom filter in some respects.

These algorithms are called "fuzzy hashing." Examples include "ssdeep," "smapsum," and "nilsimsa."

Is this what I know as 'image hash'?

> I knew a lawyer who had to defend someone in this type of case. He was allowed to view the evidence (he chose not to), but only him. No paralegals, no one else in his office that wasn't directly sitting on the defense table.

> If you're not granted specific authorization, even looking at this content is illegal, so Microsoft, Facebook, Apple, Google, etc have a very limited staff of people who are even authorized to handle this stuff.

With the risk of offending folks, this is such a retarded law/process. Holy fuck. This is dystopian-level letter of the law vs spirit of the law.

Can you suggest an alternative?

Not make information illegal to possess?

We don't make footage of car hackings illegal to own, why should this be any different?

At the end of the day, the harm is already done, no point causing more.

Reply on news.ycombinator.com