6 ms·

EFF: Americans may not realize the impacts of face recognition technology

85 points by andreyf 14y ago

sit12 14y ago
How will new face recognition technology discriminate between identical twins?
- lumberjack 14y ago
  Good point. I hope this doesn't become yet another forensic pseudoscience tool were procedures that are very much prone to error, are treated as infallible simple because they are use technology which the common man doesn't understands.
- TeMPOraL 14y ago
  It probably won't. However, with the numbers of cameras that are are being deployed everywhere, soon you'll only have to "tag" them once, and then you'll be able to track them as they move around the world.
  - stfu 14y ago
    Or you might just "enrich" the data with some behavioral information. Maybe one twin favors vanilla and the other strawberry ice cream.
- gcr 14y ago
  Context. Their preferences, friends they hang out with, brands they identify with, writing style, etc. will probably be different. These are things that Facebook can see.
- apu 14y ago
  That's far into the realm of sci-fi for now, given that basic face recognition is still quite poor. See my other comment for details: http://news.ycombinator.com/item?id=4275921 http://news.ycombinator.com/item?id=4275921
uiri 14y ago
At least for Facebook, I think it is possible to check if they have enough photos tagged of you that they feel confident that they can identify you in photos. 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.
- joering2 14y ago
  I think its worth pointing out that regardless of how you setup your photo recognition/face identification settings, what happens behind the scene is entirely different thing. 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.
  - orijing 14y ago
    That makes me wonder two things: 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?
    - joering2 14y ago
      1. No, its not bad. I was not looking at the issue from "right" or "Wrong" perspective. This is only one example. I am sure there are others. 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.
citizens 14y ago
Does anyone else here fear the combination of facial recognition + domestic drones?
mvzink 14y ago
Not surprising; most don't even realize the impacts of basic web technology. But indeed, the obvious fact that Americans have far less privacy than they realize is interesting and important. 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. [0] http://www.davidbrin.com/transparentsociety1.html http://www.davidbrin.com/transparentsociety1.html
jiggy2011 14y ago
Between stuff like google glass,facial recognition, GPS and advanced ML , NLP techniques pretty much means the end of privacy. Even for people who don't use social networks so much. 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.
- true_religion 14y ago
  I question how much privacy really did exist before the modern era of large cities and impersonal lives. 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?
  - jiggy2011 14y ago
    Perhaps, but in past times you would live your life in a small community of people who would "know" you and would know all kinds of both positive and negative things about 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.
daemon13 14y ago
BTW, how good is FR technology these days, especially in edge cases (blurred face, angles, etc)? Also, any pointers to algos/open source tools?
- apu 14y ago
  Terrible, unless you're in a "cooperative" setting, i.e., where the user is in controlled lighting conditions and looks directly at the camera, in a decent-to-high res photo. 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 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.
16s 14y ago
I've built OpenCV for use on HPC clusters, and while not a big facial coder, I have help those who do it. It's interesting to distinguish between the terms 'facial detection' and 'facial recognition'. The former basically judges whether or not a face is in a picture or video while the latter attempts to associate faces with names/identities. Two very different things. I just wanted to point that out. OpenCV is good at detection, but I'm told not so good at recognition.
burntwater 14y ago
While this has been in my consciousness, I can't say I ever thought too seriously about it. Until yesterday, when I was lounging on a lawn in the park. I watched a woman take a photo, of which I was most likely in the background. The photo will likely be posted online, along with date and possibly GPS information encoded. 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.

85 points by andreyf 14y ago

sit12 14y ago

How will new face recognition technology discriminate between identical twins?

lumberjack 14y ago
Good point. I hope this doesn't become yet another forensic pseudoscience tool were procedures that are very much prone to error, are treated as infallible simple because they are use technology which the common man doesn't understands.
TeMPOraL 14y ago
It probably won't. However, with the numbers of cameras that are are being deployed everywhere, soon you'll only have to "tag" them once, and then you'll be able to track them as they move around the world.
- stfu 14y ago
  Or you might just "enrich" the data with some behavioral information. Maybe one twin favors vanilla and the other strawberry ice cream.
gcr 14y ago
Context. Their preferences, friends they hang out with, brands they identify with, writing style, etc. will probably be different. These are things that Facebook can see.
apu 14y ago
That's far into the realm of sci-fi for now, given that basic face recognition is still quite poor. See my other comment for details: http://news.ycombinator.com/item?id=4275921 http://news.ycombinator.com/item?id=4275921

uiri 14y ago

At least for Facebook, I think it is possible to check if they have enough photos tagged of you that they feel confident that they can identify you in photos. 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.

joering2 14y ago
I think its worth pointing out that regardless of how you setup your photo recognition/face identification settings, what happens behind the scene is entirely different thing. 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.
- orijing 14y ago
  That makes me wonder two things: 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?
  - joering2 14y ago
    1. No, its not bad. I was not looking at the issue from "right" or "Wrong" perspective. This is only one example. I am sure there are others. 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.

citizens 14y ago

Does anyone else here fear the combination of facial recognition + domestic drones?

mvzink 14y ago

Not surprising; most don't even realize the impacts of basic web technology. But indeed, the obvious fact that Americans have far less privacy than they realize is interesting and important. 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. [0] http://www.davidbrin.com/transparentsociety1.html http://www.davidbrin.com/transparentsociety1.html

jiggy2011 14y ago

Between stuff like google glass,facial recognition, GPS and advanced ML , NLP techniques pretty much means the end of privacy. Even for people who don't use social networks so much. 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.

true_religion 14y ago
I question how much privacy really did exist before the modern era of large cities and impersonal lives. 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?
- jiggy2011 14y ago
  Perhaps, but in past times you would live your life in a small community of people who would "know" you and would know all kinds of both positive and negative things about 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.

daemon13 14y ago

BTW, how good is FR technology these days, especially in edge cases (blurred face, angles, etc)? Also, any pointers to algos/open source tools?

apu 14y ago
Terrible, unless you're in a "cooperative" setting, i.e., where the user is in controlled lighting conditions and looks directly at the camera, in a decent-to-high res photo. 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 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.

16s 14y ago

I've built OpenCV for use on HPC clusters, and while not a big facial coder, I have help those who do it. It's interesting to distinguish between the terms 'facial detection' and 'facial recognition'. The former basically judges whether or not a face is in a picture or video while the latter attempts to associate faces with names/identities. Two very different things. I just wanted to point that out. OpenCV is good at detection, but I'm told not so good at recognition.

burntwater 14y ago

While this has been in my consciousness, I can't say I ever thought too seriously about it. Until yesterday, when I was lounging on a lawn in the park. I watched a woman take a photo, of which I was most likely in the background. The photo will likely be posted online, along with date and possibly GPS information encoded. 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.