Hacker News Reader

Comment by bastawhiz | Hacker News Reader

What is the proposed recourse? Ban Facebook from analyzing data that users give it? I use Facebook's automatic tagging because I find it useful. Should it be illegal for me to use this feature because someone might unintentionally be identified in a photo that I took? Is the argument that Facebook--or any other company--should never be allowed to develop software that could process data that could identify folks, even when the data was willingly provided by a user that's complying with the law?

I'm not arguing in defense of Facebook, I just don't understand what someone that's privacy-minded would want. If I can legally take a photo, should I not have the right to have a company analyze that photo on my behalf?

View on HN

Not saying I agree, but an argument could be made that other people need to be blurred before it is shared with a third party, if they have a legal right to control images of themselves.

If you accept that people do have this right the logical conclusion is then you are infringing and causing harm to them by taking their photo and sharing it with Facebook. If you don't think people should have that right, I am not convinced there is any infringement going on and this creepy database is just the inevitable end result.

Symmetric key encrypt the image, upload the encrypted image with the symmetric key encrypted by your friends' pubkeys.

When you add a friend just append new pubkey encrypted symmetric key for each asset. Friend removal is even more expensive and requires the client side to reencrypt everything :(

I'm sure there are better cryptographic constructs out there with multi key encryption that can alleviate some of the insertion and removal costs.

You could do it with decent trade-offs for the content provider as well. You could obfuscate everyone else's face but yours and allow them to get access to that image and then just encrypt the map for your friends. Of course then it's all about just creating enough mock accounts to recover the full images throughout the graph. Or just take custody of a user's keys in exchange for features.

Still though, it's marginally more expensive to build a mass surveillance app!

I will never disagree with advancing software, because at the end of the day if we don't, someone else will. But I do disagree with companies taking advantage of the laws. There is no privacy law at the moment that bans Facebook from seeing you in someone else's picture and locating where you were, and targeting ads at you based on that information. But it is common decency to have some respect, and discard that information. This is where the capitalist mindset takes over and makes companies think this behavior is ok.

Devil's advocate: there's also no evidence that Facebook is spotting users in airport selfies and targeting ads based on that data.

Even if there were laws, it would be impossible to show whether or not they are (or aren't) complying.

I guess my point is, how would such a law be crafted to be a) reasonably enforceable and b) not impede on the rights of users who want such services

True but you don't train a neural network out of thin air; the data has to come from somewhere. I expect it would require radically new laws that target AI.

Blur out people without explicit consent. Draconian, yes, enforceable, certainly.

> What is the proposed recourse? Ban Facebook from analyzing data that users give it?

I think the answer's obvious -- replace Facebook with an entity, hypothetical at present, that's controlled by its users, not a corporate board of directors whose loyalty is to themselves and to stockholders, people for whom subscribers are a product being offered up for sale.

I'm obviously speaking hypothetically -- there's essentially zero chance for such a thing, given the modern Internet and the current political environment.

> If I can legally take a photo, should I not have the right to have a company analyze that photo on my behalf?

The unasked question is whether the analysis is carried out on your behalf.

Communism would't work, without some benevolent AI running it. People incharge of managing the system will have reason in billions to be corrupt. Such useful info, will always be monetised, right now we can follow the money, take away the legal way and a backdoor will open which will be more difficult to tame.

> Communism would't work ...

That's an example of a false choice. It's not a choice between corporate ownership of our information on one hand, and a failed governmental system on the other (which Communism certainly is). People being in control over their information isn't the same as eliminating capitalism. Those issues are orthogonal -- one can have freedom of opportunity alongside respect for personal privacy.

As an example, consider medical information and the rules for its dissemination. No imaginable corporate entity is so potentially rapacious as a hospital, yet the rules for dissemination of medical information are very strict, such that to reveal patient information to a third party without the patient's consent is grounds for immediate dismissal.

>Communism would't work, without some benevolent AI running it

Oh really? And who would build that "benevolent" AI? A lot of things that may be benevolent for you may not be so benevolent for others.

Facebook can of course do whatever it wants, and users as well. But bystanders should be able preserve their privacy if they so wish. The only way to do this is active obfuscation.

Here's hoping 2017 is the year it gets scaled up, because it would sure be profitable, with the right implementation.

A friend of mine had something similar[0] on his car (it was a Lamborghini) to avoid his car being identified in images online. I'm sure it's not hard to have hoodies and tee shirts with this integrated. It's not visible to the human eye so unless you hold up a camera, it'll not be a problem. That said, this might be an issue if you visit a bank or airport or something along those lines. [0] - https://www.youtube.com/watch?v=Z2oYReMaS4A

Two remedies I've heard about in science fiction are:

1) "Gevulot" in the book Quantum Thief. There are laws that allow you to control the spread of information about you. Digital technology without the 'DRM' that enforces these laws is banned (and yes, the book does explore the obvious problems with this). I've written more on this at http://www.bayleshanks.com/notes-computer-gevulot

2) "Friends of Privacy" in the book Rainbow's End. A charitable organization creates a zillion fake identities with your name so that someone getting a credit report for you, looking for you on Facebook, etc, gets a lot of hits, all indistinguishable from you, and all filled with lies.

Well in the EU Facebook can't do the kind of face recognition described in the article.

Most people in Europe actually still think face recognition is useless (like how speech recognition was useless until very recently) because they haven't experienced it on Facebook.

> Should it be illegal for me to use this feature because someone might unintentionally be identified in a photo that I took?

Regulating against it would be reasonable. What's the utility to the end user?

I don't think the article was advocating any specific course of action, but it's good to be aware of the scope of the information they have available in any case.

The alternative is that when you share something with your friends, it goes from you to your friends without being decrypted by anyone in the middle.

> If I can legally take a photo, should I not have the right to have a company analyze that photo on my behalf?

They have algorithms analyzing your photo. The solution is to run the algorithms on your computer.

> They have algorithms analyzing your photo. The solution is to run the algorithms on your computer.

Doesn't that defeat the purpose? The whole reason Facebook is able to do what they do is that they have existing tagged photos that I may not have access to. I don't need to train Facebook to know what my friends look like. If I have to do extra work to tag my photos, I'd might as well be manually tagging them.

Except that you do have a bunch of tagged photos with your friends in them, because they're your friends.

And the idea that someone who I haven't shared my photos with can still use them to tag me in other photos is not something that I would describe as a feature.

FB is a commercial privately owned publishing platform. In some countries, you need a model release to publish photos of me, in others you need a release for my kids, in others none. This is the thorny unresolved question of what jurisdiction 'cyberspace' operates under.

If you are storing any PID of EU citizens, regardless of your jurisdiction, that will come under GDPR very soon (not sure how that impacts social publishing, photos, info gathered in public vs private places, or if it has a get-out clause etc but mentioning it as an example)

Reply on news.ycombinator.com