"No one could possibly monitor that" is the same argument the pro-creepshot people use to explain why you can't ban pictures of children or candid pictures. Apparently because they think the only way to solve that is to hire a PI to painstakingly investigate each and every submission.

Those are both nuts. You handle the issue the way adults handle the issue:

1. Say "you can't post X."

2. Count on the fact that most people won't post X once you tell them not to (on pains of account sanction, which are already in place for rules against things like spam).

3. Deal with the complains of false positives and false negatives on a case-by-case basis.

> What if those photos do leak information that could enable real-life harassement

Then they should be removed. Imgur (Reddit's favourite image hosting) already strips meta data.

EXIF metadata is easy to automate. What if there is a logo of a local business reflected on a surface that most people barely notice. Someone recognizes this, tracks the person down. It's a bit over the top, but these things happen. If that exposes one of those underage girls, who is accountable? Leaking real world information can be done in many and subtle ways. There are times where people think the image origin is untraceable, but it isn't. Will reddit take the heat? The poster? The mod?

It's worse than that. Facebook already does face recognition, Google supposedly could too.

If there's a face in that image, then there is the identity of a person in that image.