EFF: Americans may not realize the impacts of face recognition technology
networkworld.com
networkworld.com
If the recognition is accurate enough , you only need 1 person to tag you once and all of your movements can be tracked by cameras (that will eventually be built into everyone's glasses).
You're out walking and you do something dumb like trip over a crack in the pavement. Some passer by taps the "that was funny, tag it" button that is mounted inconspicuously on their person and it's instantly fed into the streams of all your friends and family.
The question is, what does this actually mean to society?
If everyone's personal life is on permanent display does it mean that we become more robot like because nobody wants to have something they may regret recorded in perpetuity or does it make people more liberal because things that used to be considered funny/outrageous/taboo simply no longer are. In the sense that if every woman went around topless, men would soon cease becoming aroused by breasts.
The question is, what does this actually
mean to society?
In our country during communism, phones were constantly wiretapped. The secret service also had countless informants that were sending reports on their neighbors on anything they deemed "suspicious". Basically anyone of your neighbors could be an informant and so everybody was watching their backs constantly.
Personally I'm not that worried about technology, because a police state can spy on its citizens just fine without any of this modern technology: they just need informants and a big militia force to read and act on those reports. And believe me when I say that this method is extremely effective ;)
Quite the contrary, with modern technology regular people can use secure communication channels that can't be wiretapped, something which wasn't possible in my country during communism.
These problems (privacy, other freedoms) are always social problems, not technological problems. Technology always provides a balance ... if face recognition will be able to identify you with high accuracy, you will also be able to detect all the photos of you that have ever been uploaded and send take-down notices or other such requests. If Facebook still identifies you, you can sue them into oblivion. If your Facebook communication is not secure, you can always have a conversation over an encrypted channel. Etc...
When we all lived within spitting distance of our families in a single common room, and everyone in town knew you from birth to death, how much different was life then from now?
Will the 'erosion' of privacy really just return us to a state that we (humanity) have previously been in?
The way we're heading, we'll have neither personal trust nor privacy. It's not JUST the lack of privacy that's a problem, it's the fact that any stranger can know everything there is to know about you. In a small community, everybody has to have good working relationships with everybody else or the community dies. On the internet, not everybody has your best interests in mind.
Star wars kid could probably cure cancer, but he'd still be "the star wars kid" to most people.
I see a policeman on TV, he's a 'cop'. Sure he may be a father, avid sci-fi junkie, and so on, but all I see is one dimension.
If Star Wars Kid cured cancer, we wouldn't forget about his light saber antics but we'd add that to our impression of him----now he's a nobel winning scientist, with an impressive intellect, who has saved potentially billions of lives..... oh and in his youth he used to geek out on Youtube.
The policeman example is interesting, what happens in a society where a policeman is trying to be taken seriously and discourage you from some activity you are doing but your glasses automatically recognise his face and say "hey, I see you are being harassed by the police, these naked pictures leaked by his ex-wife might be of use to you".
What you are looking at in the future is the possibility of having people all over the world who don't really know you at all making judgements about you based on incomplete (even if still comprehensive) and possibly out of context information.
For example "cyber bullying" where what would once be bullying from perhaps a small group of people at a school can be escalated into cases of kids having thousands of hateful messages from people who they have never met posted on their facebook wall or their pictures turned into memes or whatever. For example the "star wars kid".
For example: Imagine going for a job interview in another country and having the interviewer say "you're that idiot who did X" because that is basically how you are known to the world despite your meaningful work in some unrelated area.
For example, I think in small towns the pariahs and those dubbed 'weird' have a harder time getting rid of their reputation and there was always far more pressure to conform than there is today.
When you become well known in a small town for some particular thing which may give you a bad reputation you have the option to move away or wait for time to fade that reputation somewhat. On the internet there will be hard evidence in existence for perpetuity.
Edit: And besides, in small knit societies the lack of privacy is reciprocal. It's not the same thing as having a small group of people who control massive data centers and have the capability to pick up on some dirt on anyone within hours of requesting it.
There's too much pressure from both sides. Companies, governments, etc, all want to know, and lots of people want to share.
I think the future will be one of (social, at least) privacy through getting lost in the huge sea of people instead of privacy through actual anonymity.
Whilst it is likely that you will get "lost in a sea of people" for the most part, it's also quite likely that you will "stick out" in one way or another and this might not be in a way you like or even in a way that is necessarily fair.
So far privacy has mostly been attributed to "some internets movements". But if everyone (incl. politicians and their families) are impacted, privacy should hopefully become a more central issue for policy makers.
By that I mean something on the scale of the entire facebook database being exposed or a gigantic number of personally identifiable google searches being released.
It may not even be something that happens due to malicious activity, it could be a company choosing to do something with data that impacts a large number of people's lives in a very direct way.
I would put money on nothing happening.
There have been large scale credit card leaks before, without any real lasting impact.
There have been "facebook stalking" murders.
What privacy violation could possibly be worse? I suppose genocide based on something in some online profile is theoretically possible, but even in countries where there is risk of state-sponsored attacks on people based on their online profiles (Egypt, Iran, Syria etc) there has been no real change in behaviour.
"People" (speaking in averages here) seem comfortable enough (or too lazy to do anything about) with companies knowing pretty much everything there is about them, even in the worst circumstances possible.
Stalking murders I imagine are pretty rare and will only affect a small number of people, even so I imagine people who know people who have been murdered because of facebook are somewhat more skeptical of it.
Having your credit card stolen isn't actually such a big deal, you can usually just cancel, get a new one and have the fraudulent charges reversed.
Having stuff like your preferences in porn or the conversations between you and the person you are cheating on your spouse shown off to the world, that stuff can hurt.
Not to mention that it would stifle all kinds of things, for example the kinect uses facial recognition to recognise different people. The public would not support this sort of legislation if it meant all their fun stuff got thrown away.
Besides people could still keep databases, just locked away in encrypted partitions.
In the case of kinect, perhaps small "friends and family" personal use databases could still be allowed. But large-scale databases (e.g. millions of faces) to recognize any person could be prohibited. This way, your xbox and whatever personal devices can recognize you for fun, but random strangers cannot.
Also, what happens if I move my "friends and family" database to a cloud provider?
You seem to assume that companies would be determined to circumvent the law. I am not convinced this is the case. For instance, I'm pretty sure Google and some other companies have actually held back certain face-based technologies simply because it's creepy, even in the absence of regulation. If US and EU law both prohibited massive-scale face recognition databases, I don't think Facebook and the like would be rushing to circumvent it. Are they really going to set up a server in Sealand or search for some elaborate loophole?
Finally, I just want to add that we won't be able to work out all the details within a single HN comment thread. Legislation in general is tricky and fraught with imperfections and tradeoffs. That doesn't mean it's not worth doing, or that it wouldn't be effective. If I had just proposed the idea that freedom of speech could be protected by law, it would be easy to find problems with that. But at the end of the day, the first amendment and so on is valuable and effective... in spite of any imperfections. So on the same note, I feel that privacy legislation, even though it's kind of a new thing with some issues, could be made to work.
And then the train of government/media abuse starts rolling: Persons-of-interest sought by local law enforcement? Registered sex offenders? People on the No-Fly list? Shouldn't these devices support wiretapping such that they can recognize people as directed by LEO on a rubberstamp warrant?
I am repeatedly reminded of Brin's "The Transparent Society" [0]. Given the breadth and depth of privacy violations that continue to progress in the US and elsewhere, it seems more and more likely that we will eventually face a choice "between privacy and freedom" as Brin predicts. Sadly, people do indeed remain unaware, and even the savvy aren't taking the issue seriously (read: radically) enough. I count myself among the lazy, but I'm happy to preach doom like the rest.
Though I'm in the background, likely blurred, I can fully visualize a system that would combine near-certain images (a security camera is 90% positive it saw me three blocks west of the park at 2pm), with the blurred image taken at 3pm in the park, 30% likely it's me, with another security image taken at 4pm 2 blocks east of the park with 85% certainty.
In other words, 10 years from now these algorithms won't just be able to follow my every move - they'll be able to go back and scan all the pictures uploaded, ever. And compile a film showing every waking hour of my life. A timeline, really.
Facebook → Privacy Settings → Timeline and Tagging Edit Settings → Who sees tag suggestions when photos that look like you are uploaded? (this is not yet available to you)
Needless to say there are very few pictures of me on Facebook and none of them are taken or uploaded by me.
I can see FB user turning all those recognition features off, but yet when FB is given a photo and asked by LE for help with finding you in FB network, FB can choose to comply with that and quickly find you in a stack of 100 billions of photos, providing LE with array of all your photos organized by time and place.
edit: also, I believe there will be more way of abusing photos recognition in the future. Just like Federalies are using Google Earth Maps to find who has a swimming pool and match this information against resident's taxes (whether they paid pool taxes or not) and acting accordingly if they haven't paid, I can see IRS tapping (or at least wanting to tap) into FB photos stream and find folks that bought luxury stuff and did not claim it tax-wise.
1. Is it bad (or unfair) for the IRS to use technology to fight tax fraud? I mean, every dollar not collected from a tax evader is an extra dollar that must be collected from a law-abiding citizen. It sucks for the person with the pool, but if it means lower taxes for everyone else, I think that would be a good thing.
2. Google Maps is public, but what obligations does Facebook have to reveal photos that are posted privately (i.e. to friends), to law enforcement?
2. Unfortunately, its not even up to the judge to determine "obligations". Companies that grow as big as Facebook tend to turn evilish. ATT is a private company with healthy profits; your phonecalls suppose to be private and you pay ATT for delivering of service, but yet you have NSA and others involved, where the Government publicly saying "yes we listen everything, we OCR your conversations looking for terrorism, we record everything in our trillion terabytes storage center". Skype is private too but its been known they are or may be listening too.
To think that profits-seeking revenue-troubled Facebook will not want to look for additional ways of making money such as selling your data to government is naive. I really can imagine, sooner or later, an official set aside budget that Govetnment will come up with that will be spend only on asking Facebook for full access to their databases. Who knows, perhaps this is they way FB will turn profitable-healthy. Guess the question would be if that, hopefully, will finally scare ppl off of using it.
Also, any pointers to algos/open source tools?
Labeled Faces in the Wild (LFW) is the de-facto standard for looking at real-world face verification ("are these two faces of the same person?"): http://vis-www.cs.umass.edu/lfw/results.html
The top performer is currently around 93% (random chance is 50%). For a very rough guide on how recognition rates would be ("who is this person?"), it's v^sqrt(N), where v is the verification rate and N is the number of different people you're trying to distinguish between. Note that it goes down very fast with N.