Chinese face-recognition technology spots wanted man at 60,000-people concert

Coming soon to a city near you (if it hasn't been deployed already). What can we do about facial recognition tech? Is there any feasible way to thwart it that doesn't involve wearing a mask or living in the woods as a hermit? The suspect, who was identified only as Mr. Ao, was wanted for unnamed “economic crimes.” Mr. Ao made the unfortunate decision of attending a concert by pop star Jacky Cheung in Nanchang, where he was identified by CCTV cameras and arrested. Ironically, he had drive 90km (56 miles) with his wife, to see the concert. The suspect was shocked that he was identified and said that “he wouldn’t have gone” to the concert if he knew what would happen (obviously). “The suspect looked completely caught by surprise when we took him away,” police officer Li Jin told state news agency Xinhua. “He didn’t think the police would be able to catch him from a crowd of 60,000 so quickly,” Mr Li, from Honggutan police station in Nanchang city, added. This unlikely capture is certainly an impressive achievement, but it also raises some concerns regarding Big Brother-like security. China is probably the world leader in terms of facial recognition systems, and it’s also a leader in applying this technology to catch suspects. The country boasts a huge surveillance network of over 170 million CCTV cameras, all of which rely on the collection of ID photos the government gathered. https://www.zmescience.com/science/domestic-science/china-face-recognition-system-16042018/

There are people who advocate weird hair and makeup styles that are supposed to make it more difficult for facial recognition software, but I think that is thwarted by the fact that everyone would know you as the guy who does that weird soykaf. Could move to a crowded city in Japan where it's socially acceptable to wear a mask all the time. I think more advanced recognition software is coming, that looks at your face, body, clothing, etc. I can't imagine a way to stop it, but I hope someone can

This was always inevitable, the real threat is mass surveillance.

simple solution guys, just chop your head off!

Some people just don't understand the dangers of indiscriminate surveillance

Having laws regulating possession and usage of such data would be nice.

From China with Love.

Just make yourself look like a hobo.

>>9900 But I already do.

>>9894 this but unironically

>>9902 https://gizmodo.com/this-simple-sticker-can-trick-neural-networks-into-thin-1821735479 Just stick such a sticker on your forehead and you are free of all surveillance

>>9903 Could work as big 80's-style earrings, or a badge on a hat or scarf, or a hair clasp.

I wonder if thick make up can deceive the neural network.

>>9911 You can, but it's not exactly inconspicuous.

>>9912 It would be if it was the in thing for fashion or a specifically large subculture. But that would be very difficult to pull off and maintain .

>>9836 >>9910 Saw recently a publication where they beat people in those Japanese face masks with a 60% success rate. Same for stuff like sunglasses or using a scarf. You likely have to 3d print realistic faces to beat the system unless we find a way to spot that too. There actually also exists a project like that with some artist who offered his face as the template.

it's easy to fool facial recognition software if you train an AI against it, and then put an image generated by that AI on your face. It's really an AI war. Let those with the best neural net win!

>>9914 Yes. The mask you're talking about is the one in the pic. They call it a "URME" mask. They get them from some company that makes custom masks. You could order any face you want if you have a 3D scan to give them. I suppose it would be possible to use a high res video game model for the purpose.

>>9836 Jesus CHRIST.

I was at the senior design expo for my school's engineering department and one of the (corporate-sponsored) projects was on an object-tracking camera system for use in automated surveillance. It really shocked me that this stuff isn't just being developed by distant high-level organizations, but even by dopey undergrad students at my low-tier public university.

>>9929 You probably weren't developing anything new, you were probably just trying to reach a cookie cutter conclusion, I had many projects like that when I was in school. Just think if they are having small time schools doing that, what are the big ones like MIT and actually R&D developing

>>9929 >>9930 I often wonder if the people doing R&D for these companies think much about the potential implications of their research. I imagine the fat paycheck is a huge motivator, but they must realize that they're just putting more power into the hands of various governments.

>>10023 I may just be a pessimist but in the grand scheme of things someone else will just develop the same thing you aviod, maybe just a couple years later or even months (at the rate tech is developing now) unfortunately today's society makes individuals feel powerless, surely you've felt your vote doesn't matter or that taking a stand on an issue only sets yourself back with only mild inconvenience to what you're trying to prevent. you'd need a widespread consensus from many people to make a real difference but each one will be concerned about the others backing out and fuarrrking everyone else over; ironically concerns cause the catastrophe they were trying to prevent

>>9893 >the real threat is mass surveillance Even this, is just a single part of the particular threat. The video surveillance system used in this concert is a police system, but it has been said that the government is going to install a similar system to all the traffic cameras to watch the pedestrians, and it has already been put under test in a few cities - it runs face-recognition on all the crossroads and identifies all walkers' ID numbers, to catch jaywalkers. They would be caught as soon as they passed the road, and the system will find the phone numbers from the database, and they would receive a SMS message immediately for a notice of paying the fine. The most funny and cyberpunk dystopia part was, during a test run, if the jaywalkers don't want to pay a fine, traffic police has given them another opinion: posting an warning of your misbehave on your social media account! The mass-deployment may be coming soon. And it is rumored that the subways would soon require a proof of identity. It may be an active one, like reading your ID card in addition to the traffic passcard, or a passive one, like a mandatory requirement of your ID when you register a traffic passcard. Not sure yet. It may be even simpler, just promote more people to use the smartphone application that uses NFC for your pass, and upgrade the app to link the identity to your smartphone, which has already been linked to one's ID. By the way there are video cameras equipped with face-recognition system in the subways of course... Actually, this is one of the earliest test runs. Plus all the news about the "social credit system". In China there is no systematic financial credit system like Equifax, or a public-accessible crime-record system like FBI's. The government is now trying to build one, but not limited to finance or crime, but encompassing a wide-range of activities and behaviors. For example, jaywalking is expected to decrease your "social credit". If you ended up with a low credit... Psycho-Pass, anyone? >>9894 >simple solution guys, just chop your head off! Now I see, Madoka Magica (especially Mami-san) and Durarara are something about mass surveillance... And if Deep Learning blackmagic is almost the modern alchemy, then the personal private information unethically obtained and used for the AI training would be the Philosopher's Stone, naturally the massive collecting and dragnet surveillance system over the entire population by the NSA would be the equivalent of the "nationwide transmutation circle", so Fullmetal Alchemist is in fact a story about Big Data... Or we may also learn about ways to avoid infoleak and perverse your anonymity by reading Death Note for some inspiration, https://www.gwern.net/Death-Note-Anonymity Putting all my brewing soykaf above aside... >>9890 >Could move to a crowded city in Japan where it's socially acceptable to wear a mask all the time. Or move to a polluted city (most cities) in China, where it's socially acceptable to war a mask all the time. Alas, air pollution is a serious issue, the picture attached is a typical heavy-pollution day in a windless day of midwinter. The situation is even far worse in the upper north of the country where there are lots of the heavy industries and loose regulations. The bad news is the government's policy is at least partially successful, and heavy pollution days in major cities have been significantly reduced, in extra the smog has became the SNAFU accepted by most people long ago. Therefore you would only find a minority of people who wear masks even in a heavy-pollution day, compared to 2013 where the pollution just broke out and everyone on the street had one mask. After all, the mask is just an temporary hack, not a longterm solution. Perhaps in the future, the identity thieves are going to be those people who managed to generate a special picture to mislead the neural network to identify one as another person and commit whatever crimes they wanted. Hackers would also do it. >>9898 >From China with Love. The bottom-line is, given the current situation, perhaps being a NEET is the only way one can minimize the exposure of oneself in the meatspace. But it's still problematic. You need to buy things, and all your online purchases and payments are still linked to your name and ID... So let's use cryptocurrency and underground marketplace (if we assume there are even vendors who are willing to sell food, day-to-day items, computers, or manga via a smart contract on the blockchain, or we can barter in a person-to-person manner on a online chat if you want a more doable example), problem solved? Nope... Don't forget that one's name and phone number (hence your ID) are still legally-required to use the package delivery service. There may be some ways to cheat this as well, but I'm not sure...Anonymous NEET Otaku Surviving Guide in China, is this hypothetical thing will come soon enough? so when...

Could sticking an always on infrared lightsource above your head work?

>>10023 Better to have it done at a university than in a private r&d lab. At least then we all benefit in some sense. Otherwise you get people like Google treating the public like their lab rats, then profiting off the rewards of our collective risk (think self driving cars). Relying on people not to make dangerous tech is impractical no matter how you look at it. Besides the paycheck there's the fact that you don't necessarily know something's going to be a risk until you do it. Our culture also tries to make science apolitical. Seems misguided until you consider the alternative where all science is openly done with a political agenda.

>>10030 Nihao, Dr. Pavel, Im Ministry of State Security

>>10030 >Perhaps in the future, the identity thieves are going to be those people who managed to generate a special picture to mislead the neural network Have you seen that paper on "adversarial patches" that you can use to fool image recognition?

>>10058 overheadbeacon=true if you could somehow infiltrate the system, you may be able to fuarrrk it up like in GitS, technology could come full circle and a practical use for deepfakes exists if it could be done real time https://www.youtube.com/watch?v=8LhI-e2B8Lg >>10061 the best confusion is no image at all, i.e. blacking out the feed, the problem with trying to confuse it is that any other number of methods can be integrqted, are you familiar with the motion tracking in the "AIsight" rogram (that It seems much of the technical information is no longer available, but has been around for over half a decade comercially) or the argus system? there is also a system that allows people to get fingerprints from long distances (IIRC over 3 meters)

>>10024 Interesting news today: 12 Google employees resigned in protest due to Google's involvement in a military drone program. Apparently 4000 more employees signed a petition urging the company to cancel the contract and vow to refuse any future military work. As you said, I doubt this will have a real quantifiable impact on the progress of their research. It is nice to see that some people of these people have some integrity though. >>10058 Potentially. I saw an article a few years back that suggested just that. It was a headband with a bunch of IR LEDs in it. I don't know how well it works in daylight, but one potential problem I see is that you would quickly attract the attention of security if you entered a place like a subway station. If it works as intended it would be perfect for a stroll down a public street though.

>>10064 The device. https://makezine.com/2008/02/20/temporarily-blind-surveil/ Problem is though that even if you cannot see IR it can still harm your eyes, or the eyes of other people around you.

>>10059 >I often wonder if the people doing R&D for these companies think much about the potential implications of their research. I imagine the fat paycheck is a huge motivator, but they must realize that they're just putting more power into the hands of various governments. Does it really make a difference? Maybe the massive conglomerate of megacorps is powerful enough in the West ("deep-state" meme?), sometimes may check and balance aggressive policies from the government for their own (or even public) good, if their interest would be affected. A good example is the role Silicon Valley played during the war on encryption. But even the NSA has a lot of "patriotic" businesses to work with. We've also got the "Hacking Team"-type. It also occurs in a daily basis that some companies do sell surveillance technologies to authoritarian regimes despite an embargo in effect. Google's military research, as a recent example, showed they start becoming fully cooperative as well. Also, in here, the economics works, more or less, as crony capitalism. Corporations are always the friends of the government or they won't be not able to survive from the market or the government. So the incentive always exists for companies to do these researches. >>10023 >Better to have it done at a university than in a private r&d The "military–industrial–academic complex". Not as a conspiracy, but surely the government can find teams in universities, and see below... >>10058 >Could sticking an always on infrared lightsource above your head work? >>10061 >Have you seen that paper on "adversarial patches" that you can use to fool image recognition? Yes, I've read the papers. For example, here is a new one from March describes the "infrared" and "impersonation" attack, an interesting read. >Invisible Mask: Practical Attacks on Face Recognition with Infrared >https://arxiv.org/pdf/1803.04683.pdf And note the authorship... >Fudan Univiersity, CUHK, IUB, Alibaba Inc. >Alibaba Inc. It really makes you think... Alibaba is the major company behind multiple "Smart City" "Big Data" projects, including the involvement of the Social Credit System. So the Big Brother is already one step ahead of us! >>10064 Back to the topic, I believe the problem of this technique is making yourself a large target of interest, pic (xkcd #1105) related, the system can simply detect attempts of jamming, cops can track your down and arrest you for "computer invasion", probably. Especially if you are going to use it for doing something questionable. And the "impersonation" attack also has some problems of stability and consistency, since the systems may not use the same model. >>10030 >Nihao, Dr. Pavel, Im Ministry of State Security "We shall meet in the place where there is no darkness", comrade.

>The Metropolitan Police used facial recognition at London's Notting Hill carnival in 2016 and 2017 and at a Remembrance Sunday event. >Its system incorrectly flagged 102 people as potential suspects and led to no arrests. >In figures given to Big Brother Watch, South Wales Police said its technology had made 2,685 "matches" between May 2017 and March 2018 - but 2,451 were false alarms. >Leicestershire Police tested facial recognition in 2015, but is no longer using it at events. Freedom of Information request shows Britain is like kicking dead whales down the beach at facial recognition http://www.bbc.co.uk/news/technology-44089161

>>10100 Just wait for the chinese to start selling their tech.

>>10100 There is big money at stake here, and China is a huge testing ground. They will be constantly working to improve their software because whoever is the most successful at doing so will get very rich very quickly.

i live in a small town in bosnia so i don't worry about stuff like that desu

>>10193 The Chinese lie constantly about their tech accomplishments. They are notorious for just making crap up in universities and papers. The U.S has a ton of open projects in machine vision that sound just as insidious, the recent kaggle contest for identifying people as terrorists. My intuition is that the tech will work well in situations like airport screenings (where it is already used.) A lot of countries compare your face to your passport already. Even in this situation it has a lot of false positives. For instance, before I renewed my passport I got flagged constantly because my hair was different. However, the training is actually a lot harder in situations where the camera can't get a consistent straight on look at someone. There is only so much number crunching you can do on real time video data, and inferring past people's pimples, hats, scarves, beards, and weight gain is harder than you think. I regularly use both of the top free facial recognition apps to try and learn more about people from the news (Google, and FaceFind.) From what I've seen just playing with both services, it doesn't actually take much more than a good hat and sunglasses to throw the things off. Supposedly the police oriented versions are better, but I don't believe it.

>>10197 I think you're smart to point this out. When you actually read the papers referenced by sensationalist news articles, they're typically a proof of concept that only really works under strict assumptions. For instance, a network that can identify you even when you are wearing a scarf might need to be tuned to your face with thousands of training photos. Even so, if there is a proof of concept now, there's a pretty good chance it'll be working at full scale in 5 years. Better to prepare while we can.

>>10197 Here's a brand new article about a western company pursuing facial recognition tech. Amazon is marketing their software directly to the police. https://www.nytimes.com/2018/05/22/technology/amazon-facial-recognition.html >>10197 On the whole, I agree with you. I don't believe this tech is 100% ready to go at the moment, but as I said, big companies and governments are very interested in it, and I believe it will improve significantly in the future.

>>10196 >tfw live in a small country town surrounded by religious fundamentalists who dont believe in technology, and the radio literally has advertisements for drone surveillance and freelance remote piloting services over fields. RFID and cloud based agriculture too, you can control the entire growth process from a computer. megacorps will sue you if their GMO seeds are in your field and then take your land HIGH TECH LO LIFE AND NOT EVEN IN THE CITY

>>10197 I was just about to post then I saw your post. I'd just like to add that when law enforcement mentions any sort of technology being vital to solving a crime, it usually means someone undercover/informant provided crucial information and still important to a future/on-going investigation.

Doubtful, likely fake news to spook people into submission. If humans can't tell chinks apart neither can AI.

How does it fare matching chink faces though?

>>12876 t. fart smella

>>12876 Like this

>>10065 >>10064 why not wear a mask? seems pointless

I really think adding a dick sticker on your cheeks would automatically disqualify your face as a face thus would not be processed by the system to match someone.

>>14380 ...already illegal in lots of places.

Umm I don't think a facial accessory like a mask will do. Since they'll update their databases when they see you then you won't be able to use that same accessory. There has to be an another way involving chemicals or biologic ones.

>>14954 masks are largely irrelevant in general, because it is now becoming easier to buy off the shelf solutions for gait and other biometrics, not to mention the ability for remote fingerprint scanning. The only people a mask may fool (if allowed) would be small time establishments. some places already use other data sourcing that includes cell phone pinging, and in some cases government IDs with embedded chips. you could always try wearing a burka if you did not want to be identified, but some of the more xenophobic places have put bans on those

>>14954 this is true. hence the value of a mask, is giving you a spare second identity.

>>14956 >>14960 a prosthetic nose is just as good and much harder to spot at a distance. Also gait recognition is easily fooled by walking with a slight limp. Use a RF blocking pocket for your phone and a seperate one for all cards. Easy.

>>9836 >China is probably the world leader in terms of facial recognition systems They look so the same they need computers to tell them apart

>>14956 Remote fingerprint scanning? I did a cursory search on this, and got fingerprint scanners for remotely unlocking other devices. Do you mean getting someone's fingerprints from a photo taken from a distance? That sounds too good to be true, but I'm interested.

First, I didn't read the thread. Second, have you guys heard of body recognition system? It could recognize you by the way you tread for example. Your posture also. Your face is secondary now. And the posture isn't the thing you could change easily. Wat do?

>>15375 I actually read the companies public service announcement, basically they invented a technology that allows a scanner to remotely detect fingerprints from IIRC 4 meters away (12 ft) but the product has since become very quiet. It was for a short period impemented as a gym membership door access tool for a proof of concept. >The person waves hand to sensor for identification verification--no stops at checkpoints asking for a ID card; no fumbling around for lost keycards. The machine is described as a touchless scanner. The fingerprint can be photographically captured with enough detail to match against a database. A selling point is that fingerprints are collected at a standoff distance so there is no need for people to touch a scanner pad, avoiding problems in accuracy compromised from dirt, grime, or oil on the finger. >The scenario is simple, in that the company places the scanner device on the door; the person sends prints to the system, which takes a snapshot when triggered and uses pattern recognition, edge detection and sharpening to scan the fingerprint, which is then compared to the database. The system is described as close to the way satellites process ground images. the company was called IDair... and lo and behold: >IDair has announced that it has appointed Kyu Hwang — an aerospace sales veteran — as the company’s newest director of sales for its line of touchless fingerprint biometric solutions. >According to the company, Hwang has previously held positions with Paragon Space Development Corporation, DARPA and the FAA. There was a warning to asians about taking selfies with the stupid 2 finger smile pose they are so fond of because it left fingers vulnerable to hi resolution cameras >>15408 Had you read the thread you would know all about that, also rocks in shoes is one method, except the system often integrates a number of other biometrics to reduce error margin (i.e. face camera + gait measuring +body posture + limb lengths + height) and with people carrying smart phones, it is essentially impossible not to corroborate who is who in a crowd.

>>15375 try "touchless fingerprinting" >The device works by projecting a series of striped lines onto a finger, in a process called structured light illumination (SLI). A 1.4 megapixel camera automatically captures images of the lines as they wrap around the finger, at roughly 1,000 pixels per inch. That’s twice as much as the resolution required for a fingerprint in the FBI’s Automatic Fingerprint Identification System (AFIS). By analyzing the way each line rises and falls, the software builds a 3-D model of the surface of the finger in less than a second, with each ridge and valley in its proper place. And unlike existing fingerprinting devices, the SLI system isn’t hampered by oily skin or a dry environment. tl:dr >1.4mp camera >2x the minimum required resolution to get a print

>>14995 What's scary is if a machine can tell the chinese apart at all it will be unbeatable at telling people apart, too. >>15408 >>15408 I kind of wish I had one of those systems on hand to test techniques. No system is infallible.

>>15422 Has is occurred to you that Chinese arr rook same to you because you're not Chinese yourself?

>>9836 This is just a hilarious example of the Chinese room problem. CICO: Chinese in, Chinese out.

>>15424 he's doing a bit

surveillance is one thing, crime prediction is the other https://mashable.com/2017/07/24/china-ai-crime-minority-report/?europe=true best i found quick, it's not what i wanted though. Saw an article before about ml that linked how your face looks to how likely you are to commit crime >literally getting grounded for your looks

>>15432 The USA has been using predictive analytics in crime for years now, not to mention there are algorithms that detect body language that indicates "immanent crime" that have already been shown to provide results (at least according to those who would say they are). you can even buy this police software. Some of it relies on OSINT from social media and geotaging your location (i.e. too close to common "crime areas" and you get pre-crimmed). http://www-01.ibm.com/software/analytics/spss/11/na/cpp/ https://www.youtube.com/watch?v=VXI-AJU5TCs

>>15434 behavior analytics make still some sense. What i mentioned is literally looking at your face and deciding if you are the criminal type or not. It's essentially a ugly meter 9000 for crime. Imagine getting permanently branded for how you look. Next comes plastic surgery to avoid looking like a criminal and improve your citizen score.

>>15411 Does this make it economically unfeasible? >>15434 Could something like that really detect whether or not you are about to commit a crime? I imagine a correlation between certain types of body language, behavior or facial expression and the probability of a person presenting those signs to commit a crime could be found. But I think the correlation would be a coincidental one. People, provided they aren't psychopaths, would be anxious/nervous/stressed in the moments leading up to their crime, and that would show in their body language etc. What they are identifying then, is not the intention of a person to commit a crime, but psychological or even physiological conditions that could or could not be due to the intention of the person to commit a crime. Am I underestimating how good this technology is?

>>15441 >Does this make it economically unfeasible? certainly not, you could hypothetically install a touchless fingerprint sensor anywhere. Putting them in stores would give you almost flawless analytic data on who buys what, or you could track people with better accuracy than is currently possible. > Could something like that really detect whether or not you are about to commit a crime? I imagine a correlation between certain types of body language, behavior or facial expression and the probability of a person presenting those signs to commit a crime could be found. But I think the correlation would be a coincidental one. IIRC those types of algorithms take into account vocal tone (if applicable) to detect elevated and agitated vocal patterns, and then combine that with biometrics from camera sensors to detect erratic and agitated non-verbal cues like waving arms, frowning or yelling, and aggressive body movements that indicate an assault/crime is about to happen. The accuracy is questionable, as someone with "resting bitch face" might be incorrectly identified as overly aggressive. Think of the system like that website where you submit a picture and it tells you the emotional state of the picture subject, except instead of telling you "happy/sad/angry/etc." it tells you the probability of imminent crime. A flawed system, but in the end it just needs to convince people that it works and it will have the same effect as if it did work, just dont mind the people who are false positives.

>>15442 >A flawed system, but in the end it just needs to convince people that it works and it will have the same effect as if it did work, just dont mind the people who are false positives. The thing is that false positives would be so off the charts that they would literally disrupt social order and make the store less profitable. What you're describing would be no more accurate than polygraph tests.

>>15443 It's just supposed to suggest suspects to cops. Mal cops patrol for example either way. If it gives them a better chance to spot suspects it will improve efficiency regardless of false positives. The system merely has to beat the cop's decision making for selecting who to follow to be beneficial.

>>15445 perhaps. Seems like a pretty big waste of moolah though. Unless you're amazon or some soykaf,.

>>15446 It's nothing special. Identical systems are currently used in recruiting. phone interviews are analyzed by similar software to evaluate candidates and push out a psychoanalysis profile just from the recording. I have no idea how well those toys work but it's used so the corporations believe in it and pay up for it.

>>10070 >https://arxiv.org/pdf/1803.04683.pdf >Given the tiny sizes of lit IR LEDs, smaller than a penny, we show that they can be easily embedded into a cap, and can also be hidden in an umbrella and possibly even hair or a wig. Once the devices are turned on, infrared dots will be projected on the strategic spots on the carrier’s face, subtly altering her facial features to induce a misclassification in an FR system. This helps the attacker to evade detection or more seriously, through adjusting the positions, sizes and strengths of the dots, to impersonate a different person to pass FR-based authentication. The attack is called invisible mask attack (IMA) in our research. Holy soykaf, this PAPER! CYBERPUNK IS fuarrrkING NOW.

>>15462 yeah, but nobody addresses the elephant in the room: going blind from IR eye damage

cameras can and do have IR filters. It's just gambling thereby. If you want to play agent you gotta make a full mask.

https://camerareadycosmetics.com/collections/sfx-makeup/foam-latex-prosthetics

>>9836 Seems like there are a few ways to avoid this. Glasses with semi-reflective lenses or other coatings, hats to conceal hair and distort one's facial profile, mask, etc. >>10030 >And it is rumored that the subways would soon require a proof of identity. It may be an active one, like reading your ID card in addition to the traffic passcard, or a passive one, like a mandatory requirement of your ID when you register a traffic passcard. Not sure yet. It may be even simpler, just promote more people to use the smartphone application that uses NFC for your pass, and upgrade the app to link the identity to your smartphone, which has already been linked to one's ID. By the way there are video cameras equipped with face-recognition system in the subways of course... Actually, this is one of the earliest test runs. The United States' STAR ID system seems to be moving in that direction. UPC codes on the back of IDs, magnetic strips, and potentially chips. I could see China doing the exact same thing, then making public transport payable only through one's ID.

>>15484 >The United States' STAR ID system seems to be moving in that direction. UPC codes on the back of IDs, magnetic strips, and potentially chips. I could see China doing the exact same thing, then making public transport payable only through one's ID. Star ID is something different, it is more as a stand in for federal identification (so you dont need a passport) for certain activities. I can't remember how many states do not comply, but you can still use all the services without starID if you have a passport or other government ID (government "common access card") many EU member states and the US have embedded chips in passports already, and germany has biometric and RFID scanners at ports of entry to confirm identities and that technology is already in the US too. starID is cannot be used for tracking in the same way as other things, it just makes confirming Identities easier (since in order to get a starID you need to present certain other forms of identification). Chinese as a culture are more open to government oversight in everyday activities, and culturally they may be more open to living in a panopticon if it makes doing everyday activities easier (westerners are willing to do the same so long as it is not "the gubmint" monitoring them which is ironic because of how malicious corporations like facebook are known to be)

>>10030 >Don't forget that one's name and phone number (hence your ID) are still legally-required to use the package delivery service. The runners from Mirror's Edge will likely become a reality so black- and grey- markets retain some degree of autonomy, quote me on this

It's dorky as hell, but the Reflectacles Phantom glasses, paired with a disposable face mask and headphones to cover the ears should circumvent facial recognition systems. Wear dedicated clothes and a rock in a shoe for extra profiling protection.

>>15734 offline couriers have and still are a very specialized trade, even armed courier services can be had if you can pay the fee, they just are not some cool parkour rebels against the oppressive evil squares in charge

Can't we fool this system? Do the cameras have IR filters? Like you can see your remote's IR beam on a smartphone cam. Maybe we can implant IR LEDs to fool em?

>>15742 the idea is to figure out multiple solutions for a variety of redundant sensors, each sensor can be generally fooled on its own, but when used in conjunction with one another, the ability to confirm identity increases significantly. Think of it was cell phone triangulation: one tower will give you a rough idea, two will give you a closer idea, but when bluetooth, GPS, and transmit/receive are all matched with phones pinging one another (iirc facebook or google holds the patent for phones locating one another passively) you can find people within a few feet of their actual location.

I for one support facial recognition and social credit systems. This will weed the scum from society.

>>16545 me 2, will be great to live in a hellscape where your every move is watched, dissent is illegal, and any deviation from the norm results in disappearance.

>>10236 You're in the midwest/west aren't you?

https://www.orlandoweekly.com/Blogs/archives/2019/04/17/orlando-and-tampa-airports-ramp-up-facial-recognition-scanning-for-international-passengers Just think of the convenience.

>>14380 Wouldn't one assume facial recognition being used in China can see past a face mask? Heaps of people wear them there. If you're talking about full-face masks, sure. But that could be detected by the camera and an alert could be sent to LEA w/ details about location. A feasible mask would have to be that of an already existing person. Someone famous perhaps? Making masks akin to Chinese government figures would have an interesting result. That's neglecting to mention the stupidity of wearing a mask outside in public. I'm willing to bet LE would see you and arrest you on the spot.

If you have a ton of extra money, space for a bio lab and some inverted microscopes, and either willing volunteers, surrogates, or are yourself female, you could conceivably create and bear clones in large batches. Perhaps as many as 8 at a time per female host. Producing such large volumes of identical twins capable of rapidly exchanging identities and passing as one another, or being interlinked via brain implant or common communications network would be an excellent counter measure to any facial recognition system or other ID system for that matter.

>>16827 If your ethics are flexible enough, you could use expendable batches of trafficked women and increase the number of clones per host with the expectation that no hosts survive. If I am not mistaken the theoretical maximum number of simultaneous live births with appropriate nutrition that can be sustained by a single female host (under good conditions, in good health, youthful and of large enough body size) would be around 12. I strongly suspect that batches of 6-8 and volunteers or bearing them personally would be more than sufficient unless you need an army and need it fast. Alternatively, you could just produce and bear one or two clones personally and that would provide substantial obfuscation to the ID and recognition system if you split your current identity between the clones and raise them as such.

>>16828 Bearing 12 at a time would necessitate Cesarean sections on every host and likely result in their death or immobilization given the amount of blood loss from the extraction, and all the open circulatory circuits from lost placental mass. Death in birthing is very likely with such large numbers of fetuses, and increases, on average, geometrically with the number of live births even given modern medical capabilities.

>>10064 In a world where facial recognition is used regularly by the law enforcement, having any kind of device that messes up with the cameras or stops you from being identified as a human by the ML model is akin to screaming "Please stop me and check my ID please" at the cops. Very likely the system will be able to understand when someone is trying to block it's view using IR or if you are wearing some of those adversarial patches that stops you from being tagged as a human, the system will have other fallback mechanisms like infrared and will alert authorities that someone is trying to circumvent the it. If someone were to subvert these surveillance techniques, best way to do it I feel like is making the computer think you are someone else.

>>15741 man $15 an hour to be an armed courier, that is not nearly enough

>>16827 >>16828 this sounds like scientific overkill... would be easier to have multiple identities in the system, so that you could be one of many identities instead of trying to find 1-1 body doubles. The US used to have a loophole where you could register for birth certificates with different names and then get numerous passports for the same person.

Well in crowded cities masks are not that uncommon. When I went out, I always wore one with a cooling glass and a cap and sure you'll stick but eventually you'll stop caring.

>>9836 How about don't break the fuarrrking law? And if a law is so dystopian that you can't *not* break it, then you should be fighting to abolish it anyway.

Here's something interesting: >The team at the University of KU Leuven in Belgium have demonstrated how, by simple using a colored photo held near the torso of a man can render him invisible to image recognition systems based on convolutional neural networks. >Convolutional Neural Networks or CNNs are a class of Deep learning networks that reduces the number of computations to be performed by creating hierarchical patterns from simpler and smaller networks. They are becoming the norm for image recognition applications and are being used in the field. In this new paper, the addition of color patches is seen to confuse the image detector YoLo(v2) by adding noise that disrupts the calculations of the CNN. The patch is not random and can be identified using the process defined in the publication. >This attack can be implemented by printing the disruptive pattern on a t-shirt making them invisible to surveillance system detection. You can read the paper [1] that outlines the generation of the adversarial patch. Image recognition camouflage that works on Google’s Inception has been documented in the past and we hope to see more such hacks in the future. Its a new world out there where you hacking is colorful as ever. [0] So possibly, a simple T-shirt with a coloured photo could confuse some algorithms. This does look like it'll be easy to mitigate against however, but until then this still looks cool. 0: https://hackaday.com/2019/05/03/the-cloak-of-invisibility-against-image-recognition/ 1: https://arxiv.org/abs/1904.08653 (PDF: https://arxiv.org/pdf/1904.08653)

>>17003 YouTube link: https://www.youtube.com/watch?v=MIbFvK2S9g8

>>17003 I talked about this an was ignored up top. >>16830 On top of that this patch was developed against a whitebox system.

>>14946 In Ohio they derezzed wearing a mask to protests. Now you can no longer go to protests without making potentially ruining your career.

How the hell is facial recognition going to work in East Asia where a lot of people wear surgical masks, and hair and eye shape aren't exactly very distinguishing?

>>17363 Are you having trouble distinguishing hairstyle or eye shape from the picture you posted?

>>17364 Between 5 people? Sure, okay. Between hundreds of millions? Ummm, try again sweaty.

>>17373 Plus, changing your hair color, wig or clothes along with a surgical mask is trivial and not suspicious at all. The only way the system could pin you down then is because of your body proportions/gait and your location metadata, but this is a bit of a different line of discussion.

>>17363 >>17374 Because it's a ML algo, not a Human doing the distinguishing. The algorithm is trained for this specific purpose and can likely pick up on extremely small details we couldn't without studying the faces for a good couple hours. There's no window for human error, because it's not human. Wigs, hair and clothes don't necessarily matter, because is facial recognition. Wigs that obscure the face asymmetrically and clothing with images may give it a harder time however.

>>17380 Spot on, just feed it a not racist database and it'll do the job, learning to distinguish with the same ease a native Asian can distinguish others.

>>17373 >>17380 people are really good at distinguishing individual faces, particularly faces of people they are around frequently(white people see mostly white faces, black people see mostly black faces, etc.). By the way they are designed and trained, neural nets can't get much better than humans. They can however crunch a lot more data than humans, in terms of actually looking at the photos.

>>17003 Only problem with this adversarial attack is that it's super specific to the ML algo at that specific moment in training. Can't really know how it works, and if that algo is trained further then it might not respond to that image the same later. So you'd need pretty intimate access to the CV algo in question, and you'd have to know that it's the same version used wherever you want to move undetected. Which wouldn't be possible for somebody who isn't an insider. Hopefully more general adversarial attacks are possible. Now here's another one that makes this soykaf so much worse: https://longreads.com/2019/06/21/nothing-kept-me-up-at-night-the-way-the-gorgon-stare-did/ It's really hard to even grasp how much different the future is going to be because of this soykaf. Since it's the water we swim in, we as people don't really appreciate or comprehend how much we depend on being able to occasionally slip through the cracks or get lost in a crowd. This kind of tech, when applied at scale, is going to find uses we haven't predicted, and change massive aspects of our lives we haven't predicted and haven't even really understood ourselves. And it's a Pandora's box. We can't undo this once it's implemented. Massive global computer vision tracking and surveillance isn't gonna be like Minority report or whatever where you have to slip away to Tijuana and get a new pair of retinas to get away from it. That wouldn't really be possible. And anybody who thinks they don't have anything to hide is a fool. You just aren't using your imagination properly.

On the one-hand, mass-surveillance can be rationalised in China with a population of approx 1.3 billion. How can you keep track of your population when there are so fuarrrking many of them? China's current national interest involves maintaining 'regime security', and this is the most pragmatic way to do so. The scary thing is seeing whether this form of surveillance and the social credit system that they are implementing spreads to other countries as well. There are multiple cases where people who have fallen below an acceptable social credit standard were unable to board trains, buses and planes. Talk about dystopia.

>>16830 Theres nothing illegal about wearing makeup and not carrying ID. yet.

>>17911 The chances that their model won't be tried by other governments is exactly zero. I would be absolutely shocked if there weren't already at least three nations in the world currently setting up their own version of it.