And welcome to News Hour extra with Ellen but it Jen's facial recognition technology once the thing of science fiction and thrillers is coming to a screen Nay you for real it's helping police around the world arrest suspects but sometimes the wrong ones it was based upon this new technology they said it had run facial recognition and I was a match and it's been absolutely catastrophic for you because you have gone from Big being an investment bank it's at tonight living in a homeless shelter essential Yeah I've been kind of blackballed from the industry. And that losing all my security licenses that I held so I just made it worse yet kind of destroyed my life we'll hear more of that interview later facial recognition offers an alternative to fingerprints and passwords but does it actually make a safer why is China in the lead on this what will it mean when private corporations get the software so lots to talk about and our panel today Claire Garvey an associate with Georgetown University's Center on privacy and technology who has coauthored a study into the use of facial recognition by law enforcement agencies in the u.s. And basically is much more prevalent than you would think Aaron Ross is professor in the Department of Computer Science and Engineering at Michigan State University and he works on the algorithms that make facial recognition systems better and more accurate and Tony Smith who used to be head of border security for the u.k. He's now with the biometrics Institute which looks at best practice use of biometrics like facial recognition and I just look we do a quick round from all of you on whether you think this is a good thing or a bad thing when you're worried about it or whether you think it will help keep people safe Clegg Alvi What's your general take bank or so or face recognition. Any other tool is exactly that it's a tool and it's a very powerful one which means that yes it can be a very helpful investigative tool it can be used to do a lot of good but without rules in place without very strict measures to make sure that it's not misused it can be used in ways that violate people's privacy that infringe on civil liberties and civil rights so what I want to see is face recognition being used Yes but under very strict rules and legislative guidance to make sure that it's not misused I roun Ross you help develop these mechanisms I guess you're bound to say it's it's good news well it has both pros and cons I think that's one reason why they want to ensure that the technology is working correctly that we are not satisfied with the accuracy is that we're seeing currently we want to be continuously improving the technology and from that perspective then the chances of having a mismatch would drastically reduce thank you and Tony Smith you've used these techniques as head of border security in the u.k. You thought about them with the biometrics institute should we be worried or should we be delighted that this is coming out I think in general terms it is the future we've got increasing numbers of people moving around the world now we used to have to manually look at their possible photographs we've now got technology that enables machinery to me to match people to their passports 99.9 percent who are genuine So I think generally it is a good thing and it is the way ahead but as other speakers have said it needs to be used responsibly where we are that gives us some idea roughly of where you will stand or and I just wanted to follow up with you you mentioned that you know accuracy is an issue so can you just give us an impression of how accurate the software is now if you know if how often the false results come out. It depends on the type of images to be a processing and so this is a very important point namely if you're working with constrained images namely images that are acquired under constraint conditions good lighting good pose good facial expression good illumination and perhaps a high resolution photograph then really the performance of face recognition would be in the high ninety's but when you start talking about images that are acquired in less constrained conditions in which the subject is non-cooperative for example then the performance can drastically drop but it also depends upon the gallery image against which this surveillance images matched against and so for example if the gallery image which is labelled I assume with the identity of the subject is also having a similar posts then it is very likely that you could get very good performance but if the gallery image is substantially different from the image that was acquired during surveillance operations then one is not guaranteed that you would receive or obtain a very high accuracy right now Claire can you just help us with what you've learned from your studies of the u.s. Law enforcement agencies not on the rights and wrongs of this but just what has struck you as surprising and what they can do now sure so I spent about 2 years examining how particularly state and local law enforcement agencies are using face recognition technology and this is far more advanced and far more pervasive than most people realize we are seeing at the at the far end the most advanced style systems we see 6 major agencies actively contracting for actively seeking the ability to do real time face recognition this is face recognition biometric identification off of the back of lives. Audio feeds from c.c.t.v. Cameras what this means is that face recognition enables a law enforcement agency to transform their c.c.t.v. Network their video surveillance network into a tool through which to identify people or groups of people as they walk in the street right now to just slow down a bit so it's a if you have a police car let's say with the right camera and also who are parked on a on a sidewalk or a pavement but you notice on the right by them and people walk by they can see on it is that my possible that that the police can see on their screen the names identity of who's walking by them so all major companies that contract with u.s. Law enforcement agencies advertise the ability to do this type of system we're not quite sure how many of these are you know ration and how well they work practically speaking on the ground as Professor Ross mentioned the the quality of the photos pretty much determines her accurate and then also how useful this type of system is so if the surveillance cameras aren't pulling in good quality images or if the images are off posed then the system isn't going to be very good at identifying the persons who law enforcement are actually looking for Yeah well I must have just been doing some possible faces of these are photos that demands on how you do them are getting ever ever strict as a white background exactly the right sort of size of the page and so on so that's why probably I can see tell you Smith nodding. Well that's right I mean I think that you know with technology now capturing the right image for the right purpose can be hugely beneficial and so as you say a passport has always been a token to cross borders with offices of want to know who you are and that that is your possible you haven't borrowed somebody else's possible I think it is widely accepted in border control now that actually the technology is much better than the human eye and if the photographs good and it takes a $1.00 to $1.00 match against the chip the biometric chip that's imbedded now in most passports and that's a much much better tool for law and for probably for the passenger as well but it does come back to the purpose of why. You're storing and checking these by metal for that reason it's pretty clear isn't it's to make sure that if you and you are the right person crossing the border when you start talking about other applications like face in the crowd technology in galleries then we're talking slightly different applications and I think that's where we need we need to joint down to what the purpose is Ok well let's just discuss one aspect of this now which is. Not just border security but use in ports and we'll get on to other uses through the course of the program but Tony is that right that this is the primary use at the moment airports you know passport checks that sort of stuff I think the familiar the biometric since 2 point of view there is a broad church of membership and they've surveyed their members say which of the 2 Well the big areas where this is being used in the 2 are certainly border airport scenarios and the other one is financial services to verify your identity their sort of pretty well neck and neck you know I'm not aware of a financial service where I've been asked. To show my face now this is more in terms of your identification and identifying who you are there's a huge debate in the financial sector in being able to access financial services whether biometrics not necessarily facial recognition technology but certainly biometrics is the longer term solution rather than password. Which are quite tricky as you know I say so so those are the 2 main sectors but I think you know my area of expertise is actually border management and yes they are now in widespread use around the world in terms of automated border controls and Kate could you give me a story of someone who was caught out and was rather surprised because her Yeah I mean we've had all kinds of attempts at spoofing these things you know criminals will not stop because technology is getting better and so we have had people who've taken a photo a photograph of themselves on an i Pad and then held the i pet in front of their face when they've approached the a gate and of course that will match directly to the chip in the document they're carrying because what you're matching is the live image so you tend to the to the image in the know so you so you need to be you can't just take people out of this altogether I don't think you know you need to watch people because people you know will play tricks Clare one thing is happening in the States maybe you can help us. With This is I was reading that in Arizona it well in one county in Arizona anyway they've got I couldn't believe this but they got all of Honduras is driving licenses uploaded to their database that's right that's amazing. It's it's interesting a couple of years ago the Maricopa County Sheriff's Department signed a contract with the Honduras government to gain access to all of 100 drivers' licenses for the purposes of face recognition to assist law enforcement this is a bit unique in the u.s. Because it crosses borders but the use of driver's license photos as a law enforcement tool these of the thieves recognition is actually becoming very common at least half of all Americans are now in a face recognition database that's accessible to law enforcement thanks to them getting a driver's license right and Professor Ross is that when you work on these things is that way the key 1st document the driver's license photo not necessarily there are many other applications for example trying to unlock once smartphone so you do have the opportunity to get images that are not necessarily derived from from passport documents or from driver's license documents and as was pointed out earlier there is indeed a wide variety of applications and in many of these applications the starting point could be there an individual explicitly and rolled into the system by but he cooperated fully interacting with the face recognition system yes Ok but did you just suggest that you can get the pictures from social media. That is true yeah that's very reliable is next I mean if you don't really know on a service I mean your page whether it's their picture or not. Exactly and therefore creating a gallery database based on information that is soley gleaned from the social media may not be the best way to intro a subject and clearly there the it is required to have an explicit enrollment process where one has ensured that the image being documented or the image being stored in a gallery in fact pertains to the specific individual Now there are applications where one has to recognize a face but there is no need to identify the individual for example there are applications where one could one has to determine if a certain individual was observed previously not really wanted to know who that individual was and so it is possible to engage in a recognition operation without necessarily gleaning the identity or other personal details of that individual right but you're talking there about a fallen tree in Rome and I can see that would work but but surely a lot of this isn't voluntary because people don't know about ourselves and yet that driving license image of a possible image or their fees or image probably is uploaded already into some database. And that's a very important point which researchers are handling and grappling with right now which is the purview of the intended scope and the intended purpose of the data that is presented so one of the things that we are doing is we are developing algorithms that can impart privacy to face images that have been acquired for a specific purpose for example we've tried modifying face images what we would call it altering face image so that it can be used for a specific purpose but not for purposes beyond that just as an example making modifications to a face image such that it can be used by a human reviewer but can't be used by a face match or or alternately making modifications to a face image such that an automated face recognition system can still operate on that face image but contacts tracked additional information such as age or or gender so it is possible to impart privacy to biometric data depending upon the context in which it was intended to be used that technology called solutions that are also forthcoming and we are in fact working a lot in this space I see so clear does that reassure you when you hear that this technology the sort of adaptations the technology that may help protect individuals that's great news there is a lot we can do the a legislation but if there are ways that we can build in privacy through technological solutions that's incredible one of our biggest privacy concerns is. Very much that information collected for one purpose is then used for purposes that it was not collected to be used for and driver's licenses are a good example of this we go into a d.m. Me in the us to get our driver's license to be able to drive. And yet now for half of of all of us that system that that photo is now being used to identify criminal suspects and most people aren't aware of that let's hear now from just Cheney's joins us he's a Wall Street Journal reporter in China is based in Beijing and just we're interested to hear your account of this because the Chinese are very much head of the game apparently on facial recognition technology and the use of it can you just give us some impression of what is happening in China in this area. That's right they're definitely a leader in this technology and I think I mentioned earlier that several around 5 or 6 police departments in the United States are exploring the use of experimenting of the use of real time facial recognition in China and we found in least 2 dozen maybe more departments that are actively using this technology and claim to have used it to apprehend dozens of suspects and then beyond that they use it in a whole bunch of other ways banking applying for loans getting money out of it he also for sort of molding social behavior so money found social behavior talked me through that one so I think this is the area where China probably really distinguishes itself from other countries you know they kind of use it as a sort of shame based way to keep people in line so several cities in China are using it at crosswalks where they have cameras set up and they identify people who cross against a red light and in some cases they flash their picture up on a nearby screen with a number of times they've been caught running lights they're also on a running track in Shanghai where they installed it to make sure that runners don't take shortcuts during competitions and even to keep people from stealing toilet paper and out of public toilets at parks How does that work it's a park in Beijing they found they had an issue of people going in and just ripping off huge amounts of toilet paper they installed a system where you have to scan your face and you get a $12.00 foot long piece of toilet paper and then if you want more toilet paper you have to wait 9 minutes and then you get another 2 foot long piece that stores your facial information to make sure that the same people aren't coming back. Because extraordinary is that there is more sort of political use of it I was reading about a case in China where some dissident was taking care to protect himself by not moving around with this sim card leaving his mobile at home and all the rest of it even a different province and got picked. As an actor aside I talked with and the other certainly seem to be using it to track people who are politically problematic from the government's perspective activists in China got very good at evading detection just long term cat and mouse game and so now the police are using facial recognition technology to track them and this particular activist told me that the police had told him that they have systems that alert them when he passes through certain intersections in his hometown really well I think you're toilet paper one is the most extraordinary application of software technology in face to face recognition technology yet let's just ask the others in their work on this what has really stood out which is coming to the end of the 1st half but just what has really stood out for them as a very surprising thing Claire Garvey when you're telling people about your work you know over a drink or a dinner what do you say is what shocks your interlocutor's. What shocks me the most is the use of face recognition technology as biometric surveillance and how that starting to impact the ability to participate in public protest in public demonstrations in free association we're seeing something happening in Russia right now just last week it was reported that there's a website using a face recognition up to track and identify protesters at anti corruption protests and what this means is that this is an identification tool but it can very very quickly become a tool for social control a tool to chill public speech and public protest Yeah Tony so if you imagine that as a few times as one of the big issues looking ahead but just give us another story that shot you know a surprised you or amaze you or made you laugh about how this this technology works when you hear a private sector instance like stores having their own. Database of photographs so when you come into a store they actually this is happening some power company specific like it's happened I know in some parts well that you can go into a store and I mean that you'll be identified is you know your record will be put up on what you bought last time what you what you're doing in there it's thought it was shoplifters it's now becoming more of a commercial proposition to them of you know have salesman for the side let's use a fuse in the latest line in inserts that we've got because you know they know that you were in there 2 weeks go by in a suit Professor Ross you're ahead of the guy because you're designing this stuff and it doesn't come to market I presume for months or years after you've you've designed some clever aspect of this so what what can you what do you amaze people with the next era Well I think just observing the improvement in face recognition accuracy over the past decade is very interesting and intriguing and it is certainly a testimony to the massive amount of research being conducted in the broader field of computer vision pattern recognition and machine learning I think I have to put this. To you Professor Russell Yes it's not a flippant remark I'm genuinely asking it do you worry that given what we know about how politicians behave our states behave how governments behave that you are you are helping create a sort of $984.00 a well Ian system do you worry about that. Well there are several different applications there recognising and individual becomes important. We talked about border control systems we talked about perhaps the duplications of identities in a driver state a base except or even trying to access one's own smartphone where there is private data already available and one wants to ensure that the right person is accessing the system so as long as we have applications where reliable person recognition is a necessity then I think there is a need to improve by matric technology and the balance of this is to ensure that the technology is doing the right thing and then as others have stated we also have to look at the users of the stick Knology as to how the technology is being used we can take a short break now if you want to come an extra b.b.c. You k. Tweet at b.b.c. And h extra get the podcast b.b.c. News Extra poke us you know what to buy there are you listening to the World Service of the b.b.c. Each day here on the b.b.c. World Service we take. Personal accounts of history is a beautiful day remember the sun is a field full of sheep and one sheep just dropped and I remember somebody saying I was a boy months ago because that seemed symbolic of women to take the place among men in the priesthood sex health religion politics. Culture there are no limits the kind of lubricant was just another. Record I thought so it wasn't no pressure that the problem was the least pressure. Going to record. Imports and my humans to everybody has. Been cutting hidden gems it's not just history that's in the textbook written down these are people talking about what it meant to them witness at the b.b.c. World Service don't call or subscribe to the podcast. Coming up on news hour extra in the next half hour we'll move away from airports and other security uses of facial recognition software and look at other applications there's already software that enables you to take a picture of the person sitting opposite you on the bus and then to see on your screen their social media profile and also how long before you can walk into a shop take a product and walk out with a shop automatically charging you on the basis of facial recognition all that after this new summary b.b.c. News with Jerry Smit the Spanish government says it will reinforce the police presence in tourist haunts Ponce and other crowded public areas following the 20 hardest to tanks in Catalonia bugs the interior minister why need nothing though said the terrorism threat assessment would not be raised to the maximum level of 5 police are still searching for the suspected driver of the van that mowed down pedestrians in Barcelona police in Finland say they're investigating Friday's stabbings in the southeastern city of talk who has a terrorist attack 2 people were killed and 8 others wounded 3 of them critically in the assaults police have identified the suspect as an 18 year old Moroccan man. At least 7 people have been stabbed in the Russian city of Seoul goot initial reports spoke of one attacker who was shot dead by police investigators say they've identified him as a local man the Lebanese army says it has launched an offensive against the last remaining Islamic state militant stronghold on its border with Syria the Army says there are about 600 militants and they're armed with anti-aircraft and armor piercing missiles as well as snipers and drones. The president of Nigeria Muhammadu Buhari is to return home later today after receiving medical treatment in London president Bihari had left Nigeria may for treatment for an unspecified illness. Japan has successfully put the 3rd of its homegrown global positioning satellites into orbit a 4th is due to be launched in the coming months the system will go live next year to give car navigation systems smartphones and many other devices positioning accuracy down to a few centimeters a safe you know has been how then Pakistan for a German born nun who devoted her life to battling leprosy route found lived in Pakistan for more than 50 years starting as a missionary in the slums of karate b.b.c. News you're listening to News Hour extra from the b.b.c. World Service with Alan Bennett chance in our panel today we've got just Chen Wall Street Journal reporter in China Professor Ross from Michigan State University Tony Smith who's now with the biometrics Institute and Claire Garvey of Georgetown University so we just want to before we go on to the laws and how we deal with this just go through some other applications that are interesting and surprising Claire you've got one called Find face tell me how that works that's right Find face is an application a mobile phone app in Russia that basically allows you to take a picture of a stranger maybe that attractive person sitting across from you on the Metro and upload that photo that system compares that to that person's social media accounts and will give you an identification so basically it allows you to identify strangers. The challenge here is Sanger's So you're on a bus and you're take a picture of the person you're not on your phone screen it will say well give it a Facebook page or something that's right right now it only works against v.k. Which is sort of a Russian equivalent of Facebook the app developers were able to scrape all this so the profile pictures from v.k. To build their database this is not yet been possible thanks to the controls put in place by Facebook so it hasn't been widely adopted outside of of Russia yet but this this raises a lot of concerns especially for people who are worried about being stalked this app has been used to docs or to publicly shame women who work in the adult film industry and now as I mentioned previously it's actually starting to be used to identify protestors at government or anti-corruption protests so we can see how maybe a a fun idea a social media application of face recognition can be used in all sorts of other ways that raise very serious privacy implications another application I don't know who can help me with this one dating apps when you want to date someone who looks like a celebrity you like the look of. I believe it's being used in the u.s. As well but it's definitely being implemented in China by one of the biggest dating sites in the country so you can you can sort of specify that the faces that you like whether it's a celebrity here or just someone else's face and then the apple will spit out matches people kind of rank them according to how closely they resemble your ideal woman. Or man. And Professor Ross as you look ahead what sort of applications should we be expecting to come new ones I think it is recognizing frequent visitors to a certain retail store to ensure that they're given more attention by the sales people because you're not observing an individual who's repeatedly entering a retail store to make a purchase and clearly here it is not identifying an individual but merely suggesting to the sales people involved that a repeat customer just walked in and just finally on this sort of section Josh of unusual applications I think you mentioned A.T.M.'s so cash machines. Does that mean that in China you can really just walk up to the a.t.m. Machine and put in the number put in the amount of money you wanted it comes out. Well it's actually not quite at that point yet and it's should be clear it's sort of experimental There are a couple of banks that have rolled these out as a few 100 of them spread around the country and basically it's words used for for another form of authentication so you can use your card. As well as your face and then maybe also a pin number so that in some cases an added layer layer of security not necessarily just a convenience. Right Tony Smith going to move on now to how you know legislation could be framed to protect people we're just about to hear this case which I'll play the moment of this this man in the United States who have the most life changing terrible experience of this. So clearly there are problems how do you do you see the way to frame laws that control the use of this so that it is beneficial rather than malevolent Yeah it's a good question I think it comes back to how do you define in Laura sensible proportion and reasonable use of data because biometrics is really just another form of data so we've had data issues for a long time when data protection legislation which really sets a framework for what are you know why you capture in this stuff are you keeping it and if you are keep me why I'm for how long and what are you going to use it for and I think the legislation needs to probe those areas to develop case law which would not be unlike data protection case law to ensure that see there are ways that come down and there are certain limitations on what is reasonable and what is not reasonable use and that isn't really there at the moment is not really there and it presumably is quite different challenges for state that's a state abuse and private sector appears Well absolutely and from country to country you see I mean we've just heard stories you know different parts of the world where there's a completely different mindset in a community or an attitude of Palestine could be used even if you look at the You compared with North America the e.u. Adopts a much more stringent test on the use of data in the exchange of data between government agencies National alone international level whereas in the u.s. You will find there's much more propensity to share data and I think that's what we see will emerge but what we're trying to do at the institute is to come up with a best practice kind of a reasonable set of standards is quite challenging when you're looking global at this you know what from what I understood the strictest rules in the u.s. And I tried a. Little at Clare you see that in the u.s. Where the police forces. Sharing a lot of data where is the u.s. In terms of rules to control this sure so as it stands now there are no federal laws and no state laws that comprehensively regulate the use of face recognition by law enforcement we do have 3 states now that have passed commercial face recognition that just nation but on the law enforcement side we have a couple laws that deal with specific applications for example the use of face recognition in conjunction with drones or in conjunction with body cameras but none that actually address in a comprehensive manner how law enforcement can and maybe more importantly can not use this technology with the police with the absence of rules as you're describing it does that mean that you know an individual police officer has a pretty free hand in this area in the United States so we surveyed about $130.00 law enforcement agencies and counting and what we found is we actually found about a quarter of all law enforcement agencies in the u.s. Can access a face recognition system and then it's a Wild West in terms of whether or not there's a policy in place and what that policy says so in some jurisdictions so in Michigan for example there actually are pretty strict rules about when law enforcement officers can use the technology and on whom whether they need probable cause or reasonable suspicion to suspect somebody of a crime before using the technology elsewhere in Florida for example law enforcement officers are actually encouraged to use face recognition or mobile biometric tools whenever practical that's what the policy says so they're actually affirmatively encouraged to News Face recognition in any sort of scenario that they feel is appropriate and just just on the cultural aspects of this one for example and try to you describe that extraordinary thing where people are basically walking across a pedestrian light when it's red. And they get shamed by having their image come up saying you know you've done this 3 times do people say that's wrong I mean if you do that loved the riot So what China would he do that right I mean this is actually one of the aspects of China that I think makes it sort of a really interesting laboratory for testing what this technology can really do and be used for because in China the threshold for privacy concerns is much higher in China people just are used to the idea of authorities and surveilling them and so you know a lot of the Chinese people we talk to about this technology that's kind of shrugged their shoulders and they were like well you know I guess it's good it catches criminals so that actually allows companies and police departments in China to apply this technology without worrying too much about sparking a backlash Let's take a look at what happens when it goes wrong I mentioned this and we heard from him rather than getting Stephen Talley who's actually suing the f.b.i. And the City of Denver after he was arrested based on surveillance video that appeared to show him robbing 2 banks on 2 separate occasions 4 months apart and he was able to prove his innocence both occasions when it actually had a rock solid alibi but the simple fact that there is apparently a bank robber out there who looks just like him he says has ruined his life and the arrests mean he lost his job he has an investment banker he now lives in a homeless center and in fact he you know he can't afford a phone we spoke to him on the sentence by phone well see the 1st time it was based upon a individual he had seen something on t.v. Called Crime spotters where they try to get the community involved they share a picture of a of a suspect and apparently because the similarities. He basically identified me as a bank robber even though it was an investment banker and actually working at the time of the robbery but that kind of got the ball rolling. But that was the 1st arrest right but that's the one in that case it was it was old fashioned I was in a sense in the fight that was put up on a t.v. Screen and someone said it was you know in the 2nd or Cajun it was more this new software I write my miners doing was they had new evidence so 13 months after the 1st incident I was rearrested and it was based upon this new technology they said that had run facial recognition and I was a match there's 2 types of facial recognition technology is one that's just entirely computerized and my understanding in my particular case was the f.b.i. Denied at the detectives request to do just a computerized mathematical model which is I guess completely objective just looks at a face as numbers and spits out you know this guy is very similar features what he did he made a request to an individual that made the comparison so as a human comparison based on the facial recognition technology and then the human comparison said that the features were similar ideas been absolutely catastrophic for you because you have gone from Big being an Investment Bank of America living in a homeless shelter century I've been blackballed from the industry. And that losing all my security licenses that I held so I just made it worse now what do you conclude from all of this about the software and the new technology and the fact that this facial recognition can be done with really no human intervention Well I think it's still kind of scary I think it's not proven I think that from what I've read and what's occurred to me that I think there's going to be a lot of people unfortunately that will get picked out of the system I think a lot of the facial recognition technology is there getting photos not only from people that might have been arrested and have a mug shot but I think from my understanding is now they are actually acquiring photos just from social media and without any. One's knowledge or you know regards to privacy just adding that to their huge database with the f.b.i. And based upon that alone you're going to start to get more and more mis hits you know more and more errors My understanding is with minorities it's even a higher percentage of erroneous Miss identifications. So yeah I think it's never going to be an issue you have to be some kind of controls or protocols you know placed on the system until they could you know prove that a little more accurate than it is now because obviously part of the worst case scenario what could happen yeah it's kind of destroyed my life and that was Steve tally that in the United States now a cloud Garvie it's a terrible story and it seems to involve both the technology and and human error so those sort of is as the mix of the technology and humans who are running it is it familiar to you that kind of thing if you have that kind of thing elsewhere so my understanding of the tally case is that it actually was a human review so this was a what they call a morphological comparison a human taking a look at the 2 photos not a machine but that said the potential consequences are very similar we have a case happening in Florida right now where somebody is on trial for selling crack cocaine to an undercover officer the officers took the person's photo a few weeks later they submitted it to face recognition and the face recognition system came up with 5 possible matches the analyst only sent one of those photos the photo of the person on trial back to the officers who said yeah that's the guy who sold us crack cocaine 3 weeks ago and they arrested him based off of that this individual maintains his innocence but he is on trial pretty much exclusively because if face recognition system said he was the most likely match or he was one of 5 likely matches you know Professor Ross The trouble is when you come up with this type of you know people trust that. And one of the ways in which unfortunate and tragic and terrible incidents like these could be avoided is when faced matchers require to output 2 kinds of math scores one indicating to some extent what is the probability that these 2 fish face images are of the same individual but also another number indicating what is the probability that these 2 face images put into 2 different individuals what we call as the likelihood ratio and if that type of statistic is provided to a human reviewer or even a human expert who is making the final decision then now that individual is not looking at a single number that states how similar the face images are but also is providing with the other side of the coin namely what are the chances that these 2 face images actually originate from 2 different faces and this is where the issue of trust comes in in the context in which you stated Namely we have to look at both these numbers and without knowing either of these it becomes rather difficult or in some cases terrible to even state with any clarity that these 2 face images put into the same individual the likelihood ratio test is a very important test to stick that needs to be further investigated and that is something that we are doing for other modalities as will even for fingerprints and for Iris to actually provide these to contrast in numbers if you will and then using these numbers in conjunction rather than only presenting one side just for that sunny Smith you have to sort of implement these kind of policies are you saying that you say to the police officer Ok there's an 80 percent chance that you've got a match that but it's a 20 percent chance it deserves. I don't think those are complementary So this is where the likelihood ratio test as discussed becomes interesting because it's not 80 plus 20 equals $100.00 right it could be that there are 2 face images that look exactly the same but actually put into 2 different individuals like identical twins whereas there could be 2 face images that looks likely different but put into the same individual just because there has been an aging process and therefore these 2 numbers are not complimentary but do provide distinct pieces of information about the 2 photographs that are being compared Tony Smith How much would it have helped you when you were in charge of Britain's big job and lots of people coming in to have numbers like this to give out to the offices. Well I think we need to see by metrics as a tool rather than a solution and I think that's what's gone wrong here it's been seen as the be all end all evidence which determine whether or not this person was guilty when actually it's really just a tool which is one of the pieces of the jigsaw and clearly that's where that one went wrong I mean I think what I would say in defense of gallery is I do happen to know that there are significant galleries of photographs good quality photographs held by international organizations such as Interpol and you repulse of criminals and terrorists and we don't necessarily know what passport they're traveling with these people can buy a 10 possible very easily by bribing somebody somewhere and so you know the question that these international law enforcement says look we want to protect the public we really want to stop these people don't we crossing borders why don't we have a biometric watch list which we upload onto the photographic databases that you're checking people against But in fact in reality very little is being use is being made of that sort of technology for law enforcement purposes quite often because we tend to use it more as a tool full facilitation and the identification of the universe then for law enforcement purposes where we could make better use of it in some scenarios and so I think we do need to find this right balance I've got a bridge I've got a gallery going to go going to be a bad guy so you're not you're not using that at the moment no not routinely it is you as a secondary school where you've got doubts about somebody but I mean apart from the most prolific offend as well you may be able to put some of those photographs out on the front line routinely by matric watch lists are not being used across the world to stop terrorist trouble which is a key one of the key principles of u.n. Security Council Resolution 2178 so I think this technology can really help us I think to deliver law and order and to use good purposes but it cannot be the be all and end all you still need the right and that analyst to determine whether this is right or wrong leg of your course. To learn that that you know Interpol's photographs of of suspects is not being used but apparently in the u.k. To do it to compare with people coming in in the States I imagine there will be more use of those kind of pictures at Borders So Homeland Security has not been terribly trance parent about how they're using various databases as they roll out what's called biometric exit which is right now where they are implementing face recognition basically on state and certain gates and eventually across the entire u.s. System people at airports when they're exit in the country will have biometrics the scan taken to determine exactly what Tony was saying that they are traveling with the that their documentation matches who's actually holding that passport my biggest concern here is that it will in fact be used for its intended purpose and not for other purposes you know it's you know but if we take Tony's comment to its logical conclusion let me put this you Teddi Smith who suggests that governments who are using this technology to pick out you know United States people protesting as an anti police protest they're using it to identify dissent. But not to identify people who actually threaten public security and I thought it was usually about government purposes and well I think it tells you something about where the line in the sand needs to be drawn in the big debate that we're having about where the line needs to be drawn for me as a former law enforcement you know lifetime law enforcement officer I could see value in trying to stop people crossing our borders inbound and outbound actually and I was very interested in talking to technology companies and people like Interpol in Europe all about whether that was feasible but it's not something traditionally that goes on right now which is name based very name but so if you're satisfied that someone's traveling in their right passport in the right name you've got much more chance of catching them this was probably in someone else's possible on a purchase passport you know using biometric technology then perhaps you're missing a trick and I'm quite clean that we would close that gap to Ok so just change just give us a word on China and what you know you probably don't know what they do exactly but if you got the impression that crossing a border in China would involve the comparison of your face against the database of wrongdoers as they've said. Well we don't know exactly what's going on in China we don't we do know they're starting to implement it. In airports. Very in a very sort of experimental way but one where area where they are using it borders in conjunction with other biometric. Technology is in sin John which is the north western region of China the Muslim the homeland of a Muslim majority. Or a mostly Muslim minority and where they've had some terrorism problems so they are using it in poor areas there and they do have people on terrorist watch lists and they claim to be they claim to be using it to to apprehend terror suspects so let's finally just broaden this out and think about some of the broad things that are coming up. In our world and this does seem an extraordinarily powerful technology with massive potential which is what the reasons we're talking about it Tony Smith as you look ahead you know with the biometric since you think about this stuff all the time is this one of the greatest threats in the middle you've said lots of good reasons to use it and everyone gets that but is this one of the greatest threats to previously. That that is out there. Potentially if misused Yes it is it has to be proportionate and reasonable use because we've heard some stories where this is clearly off the wall by a metric some folks gossiping captured and used for a purpose completely irrelevant to the purpose of capture in 1st place and no real policy agreement behind it but I do think it has huge potential. And you know when you look at things like the global movement of people the fact that you know we're going to get more people moving across borders around the world more and more having a passport versus having digital photographs that will lock your identity and it stops working hugely beneficial to arsons travelers in the future so be able to perhaps walk through the airports we always only cameras you know you have to stop so there you go you know what will happen next and several hack into the system and change all the photographs you know that's that's the human experience is not as if you get one level of technology is of a base I don't know I mean I think the technology is moving far forward far far more quickly than we realize it is it's very very good and getting better all there is a huge amount of global investment and I think we need to embrace it and see how it can miss best benefits and Claire Garvey and to previously a challenge to privacy definitely a challenge to privacy I would echo everything that Tony Smith said in terms of the potential benefits of this technology and how quickly it's advancing but however quickly it's advancing the the attractiveness of more advanced applications of the technology its use for wholesale surveillance is used to identify strangers on the Metro. Those are also going to become a reality so absent the the policy debates absent legislation we really are going to have a problem when it comes to you to discounting privacy and shortchanging privacy and civil liberties at the sake of convenience or security and just turn from what you've described when it comes to the challenge to privacy. H.-i there is . At the forefront of that particular field. Right yeah I mean I think it's I think in China it's full speed ahead they're not they're not worried so much about about security at this point and companies are raising huge amounts of money we just my colleague just reported on one of the major fatal facial recognition companies in China just getting of another $400000000.00 in funding in and sort of climbing into into unicorn territory as a startup. It will be interesting to see. As this goes further whether whether people in China do start to push back against it because the uses of it are really quite extensive Professor Ross just finally I wonder if you can help us with the dilemma that you must face as a scientist and scientists have often face this dilemma in the past that you know you come up with some technological breakthrough and then there's the question of how it's used in the temptation of the sciences is to say that's for the politicians to decide is that good enough or do you with your knowledge of this need to get involved in the debate about what legislation should be introduced. One of the things we are actively involved in is trying to see how the technology that we have to signing and developing and implementing can take into account the a cut concerns that I've expressed about the technology so we do not necessarily have to see biometrics and privacy being at odds with each other one of our more recent research activity is trying to ensure that the recognition algorithms that we're developing can take into account some aspects of privacy thank you very much indeed and thank you to our other panelists as well so that's Justin professor and Ross we just heard from the Tony Smith and Claire Garvey and that's it for this week's news extra comments news out extra p.b.c. Dot co dot u.k. Tweet at b.b.c. And h. Extra the podcast The b.b.c. News hour extra poke us get it subscribed on your system it'll come to a device once a week and our discussion on a single topic every week but for now that's it thanks very much for listening from I am but a chance in London. Before the knees a voice that needs no introduction. It's 100 years since the 1st of. The Fitzgerald and Dizzy Gillespie. Problems celebrate some of the greatest. The b.b.c. Promised as p.c. World Service dot com. And in an hour this month's global beats is celebrating the diversity in vibrancy of Scotland's burgeoning music scene global beats from Edinburgh with Vic Galloway with an eclectic lineup of new challenge spanning the musical spectrum and mixing Scottish sensibilities with global entrances that's after news hour on the b.b.c. World Service the world's radio station. Welcome to News of from the b.b.c. World Service.