Hacker News new | ask | show | jobs
by perfmode 2702 days ago
Was it a bit for bit copy of a circulated image? Or a screenshot?

Because it’s possible to ship a bloom filter to detect banned images without breaking e2e encryption

3 comments

You can also block screenshots, scaled, noises, recolored, cropped, occluded images, but you have to send a larger model to the device.

That’s not the point though. The point is that the expectation moving forward with these platforms is that you police yourself, or the machine learning algorithms will do it for you (and also ban you from mortgages, airplanes, employment, etc, if it has to intervene too frequently).

Facebook does (in a very limited way) this by upranking posts it thinks will boost engagement.

LinkedIn is a bit further down the slippery slope, since it’s able to influence hiring decisions (and is joined to the MS Office 365 cloud, presumably).

Amazon Now had some related screwup recently where it redlined minority neighborhoods in a bunch of cities, since it used spending habits to predict they’d be less profitable.

> Facebook does (in a very limited way) this by upranking posts it thinks will boost engagement.

Facebook (and Twitter, and Patreon, and countless other firms) will block you from posting if you post something that they regard as violating their policy. And "what is against policy" is a fuzzy enough line that "don't use Winnie the Poo pictures to make fun of the effin' President of the country" is a lot clearer than that. You may agree with the policy or not, you may think it's misguided, but WeChat is a private firm, and private firms aren't restricted by anything like the First Amendment. They provide services to you at their pleasure.

If e2e encryption is on one extreme, wechat is on the other. You have to link your account to a government id if you are the owner of a large group chat.
I’m pretty sure there’s no e2e encryption in WeChat. I communicate in WeChat with the assumption that everything I write can be read by censors.