Skip to main content

tv   House Oversight Hearing on Facial Recognition Technology  CSPAN  June 5, 2019 2:31am-5:16am EDT

2:31 am
2:32 am
this hearing begins with opening statements from the witnesses. >> thank you director. you are now recognize to give your statement for five minutes spent thank you chairman cummings and ranking member jordan and members of the committee. i am the deputy assistant director leading information services branch with the fbi criminal justice information services division. thank you for the opportunity to appear before the committee to testify today with the fbi use a facial recognition for law enforcement purposes it is crucial authorize numbers of the communities have access to today's biometric technology to investigate and identify to prosecute terrorists and
2:33 am
criminals. the fbi next-generation identification system which includes facial recognition aids in our ability across the country facial recognition is an investigative tool to greatly enhance law enforcement capabilities to protect public safety. as the fbi it is crucial to protect the civil liberties of the american people as part of our culture this is why when the fbi developed facial recognition technology and also pioneered best practices to effectively deploy these technologies for public safety to keep with the law and without interfering in the fundamental rights the fbi division has two separate programs using facial recognition technology that is the interstate photosystem or facial analysis comparison and evaluation or face services unit. typically allowing authorized law enforcement agencies the ability to use investigative
2:34 am
tools a facial recognition by looking at criminal mugshots photo lineup seven like to for decades while this practice is not new the efficiency of the searches has significantly improved using automated facial recognition the fbi policy and procedures emphasize the candidates return are not positive identification the searches are photos that only result in a ranking of candidates the fbi requires them to follow the nti implementation guide and facial identification working group standards for performing facial recognition comparisons placing legal training and security requirements on law enforcement users including a prohibition against submitting photos obtained without with the first or fourth amendment those photos in the repository
2:35 am
are criminal mugshots acquired by law enforcement partners with criminal fingerprints associated with it the fbi face services unit provides investigative lead support the offices operational divisions and legal attaches having trained to compare those images of people associated with open assessments or active investigations against facial images state and federal facial recognition systems to establish agreements with state and federal authorities. it only searches photos that have been collected pursuant to the attorney general guidelines as part as an authorized fbi investigation and not retained. the service does not provide services but investigative lead since the review of the last oversight hearing in 2017
2:36 am
the fbi has taken significant steps to advance the technology. at the end of 2017 the fbi validated the accuracy rate and in early 2018 the fbi require law enforcement users to have completed facial recognition training with the face standards prior to conducting facial recognition searches. additionally the fbi collaborated to perform the facial recognition test to determine the most viable option to upgrade the current algorithm. the algorithm chosen listed and accuracy rate 99.12 percent to leverage it is complementing the upper rate that racial effect recognition algorithm i like to thank the men and women for their unwavering commitment of the fbi so much staff
2:37 am
protecting the country against horrific crimes also the members of this committee for their engagement on this issue on behalf of the american people and law enforcement partners. thank you. >> thank you very much. >> chairman cummings ranking member and members of the committees i am pleased to be here today to discuss the work on the fbi's use of facial recognition technology over the past few decades this technology has advanced rather quickly and now has wide reaching usage from smart phone to social media and helping law enforcement and criminal investigations. however the questions we exist regarding the accuracy and the transparency of its usage and protection of privacy and civil liberty and when that technology is used to identify people based on certain characteristics. today i will discuss the
2:38 am
extent to which the fbi has ensured adherence to laws and policies related to privacy and transparency regarding the use of facial recognition technology as well as the fbi has ensure the capabilities are sufficiently accurate. will also provide uppity updates on the priority recommendations in april of this year regarding this technology. and in may 2016 report we noted to legally required documents the privacy impact assessment and the system of record was not being public in a timely manner these documents are vitally important to privacy and transparency because they analyze how it personal information is collected stored and managed while the thorn talks about the very existence of the system and the types of data collected among others.
2:39 am
doj has taken action to expedite the development process of one that has not updated the process of issuing the thorn. we also record reported on accuracy concerns of the capabilities specifically we found the fbi conducted limited assessment of the accuracy of the facial recognition searches before they deployed the technology. facial recognition system generates a list of the requested number of photos the fbi only assesses accuracy when users request a list of 50 possible matches others could have given different results for additionally these tests did not specify how often incorrect mast matches were returned knowing all of this the fda at one - - fbi built technology often uses
2:40 am
facial recognition systems operated by 21 states into federal partners to enhance the criminal investigation. we reported the fbi had not assess the accuracy of these external systems and as a result we don't know how accurate these systems are but yet the fbi keeps using them. moreover, we found the fbi did not conduct regular reviews to determine whether the searches were meeting users needs. we made recommendations to address all of these accuracy concerns doj has yet to implement them. as you are aware in april of this year the priority recommendations report which provides an overall status to outline those that gao believes it should be given high priority. this report included six recommendations so as of today
2:41 am
five of those remain open. the use of facial recognition technology raises concerns about the effectiveness of the technology to aid law enforcement and the protection of privacy and individual civil liberty. this technology is not going away and it's only going to grow brick what is important doj take steps to ensure that transparency of the system so the public is kept informed of how personal information is used in protected and the implementation of this technology protects individuals privacy and the technology is used or accurate and used appropriately. chairman cummings and ranking member and members of the committee this concludes my remarks but i'm happy to answer any questions you may have. >> thank you very much doctor. smacked linking members i'm from the information
2:42 am
technology laboratory at the national institute of standards of technology. thank you for the opportunity to appear before you today with our role in the standard of testing for facial recognition technology. in the area of biometrics nist working with public and private sectors since the 19 sixties echo improving the accuracy usability interoperability and consistency of identity management systems to ensure that the united states interest are represented in the international arena. research has provided state-of-the-art technology benchmarks and guidance to industry and us government agencies that depend upon biometrics recognition leading national and international consensus standards using biometrics such as facial recognition but also also cryptography to secure network protocols, reliability and
2:43 am
performance testing. essential to accelerate the deployment and communication system that is interoperable reliable secure and useful. this evaluation reports gaps and limitations of current record recognition technology with the advance measurement by providing a scientific basis with consensus -based standards to have that data for this purpose standards. conducting the face recognition grand challenge to challenge that facial recognition community to break new ground solving research problems on the biometric frontier since 2000 the vendor
2:44 am
recognition program look at that algorithms for one too many identification and verification. to expand the facial recognition valuation and 2017 broadens the scope of its work to understand the upper limits of human capabilities and how these fit into facial recognition application. this biometrics research assist the federal bureau of investigation and the department of homeland security. with a former us visit program. and then to analyze facial recognition capability with image quality and demographics
2:45 am
with those match algorithms those thresholds and that vendor testing program to provide independent evaluations of the prototypes with those algorithms. significant progress made in those improvements since it was created. researching how to have the examiner's different photographs and to have face edification accuracy for a group of professionals to work under circumstances approximating real-world casework. the findings of the national academy of sciences show consumers and other specialist including currents and forensic trained and untrained recognizers were more accurate than the control groups on a
2:46 am
challenging test it also presented data comparing state-of-the-art facial recognition algorithms with the bay best identifiers. optimal recognition was achieved only when humans and machines collaborated. and rigorous testing standards can increase efficiency in government and industry to expand innovation and broaden opportunities for international trade and conserve resources to provide consumer benefits and improve the environment health and safety thank you for the opportunity to testify on facial recognition i'm happy to answer any questions you may have. . >> and morning chairman cummings.
2:47 am
without transportation security administration with that analysis went to thank the committee for working with tsa as we continue to improve the security of transportation system and for your support of officers and airports nationwide. to charge the and industry was security and a key component is to positively identify passengers boarding aircraft so with that appropriate level of screening that present themselves to a security officer since its inception it is carrying out the role as efficiently as possible using available technology to recognize the need to identify in an era where they become more sophisticated and prevalent to improve import
2:48 am
performance so the biometrics roadmap that it takes to test and expand the capability at the tsa checkpoints that can enhance security and improve passenger experience. the roadmap is for major goals customs and border protection and biometrics for individual travelers. and then the pre- check passengers and for the additional domestic travelers to develop the infrastructure of these efforts. tsa conducts pilots to look at identify that certain airports looking at the applicability of tsa operations.
2:49 am
to be executed in conjunction that it has been a privacy impact assessment and passengers always have the opportunity to not participate. in these cases standard identification process is used. every passenger chose to use that biometric indication process the facial capture camera was an active mode needing only capture the image after he was in position and the officer activated it then passengers moved rapidly through the checkpoint in that regard if it gives a unique opportunity for tsa this can increase the effectiveness of the entire system using biometric identification and increasing the checkpoint enhancing passengers experience while providing
2:50 am
more effective identification will be beneficial as we see increasing passenger volumes growing at a rate of 4 percent annually. the fact we experience the business travel day ever ahead of memorial day weekend screening approximately two.8 million passengers tsa is committed to addressing cybersecurity concerns and in that regard the tsa modernization act the assessments of cbp developed with dhs science technology directorate addressing accuracy error rates and privacy issues associated with biometric identification looking ahead tsa plans to build upon the success looking at select locations with the
2:51 am
biometric implementations at the checkpoints by privacy impact assessments to clearly identify through airport signage and passengers always have the opportunity to choose not to participate. tsa is in the process of a systematic assessment this process will enhance aviation security to make air travel a more enjoyable experience using for passenger identification to determine the appropriate level of screening not for law enforcement purposes and the opportunity is always there not to participate thank you for the opportunity to address this issue before the committee i look forward to answering your questions. >> and now recognize myself. and 2017 the gao testified
2:52 am
have signed contracts with at least 16 states to request searches of the database. gao also with my chat with corrections photos. on how that database how it searches its own system and how this policy is determined? . >> i would be happy to explain that we have a service safe services unit with the background checks and facial recognition services of the state dmv photos in accordance
2:53 am
with the attorney general guidelines and then to have the open assessment or active investigation and then a search to the state the state runs the search for the fbi and a candidate list comes back. and with the interstate photosystem it will utilize that repository as well as dmv photos however state local and federal agencies only have access to the interstate photosystem and these are the mugshots associated with the cards with a criminal arrest record. . >> and those in the non- criminal database to have them
2:54 am
searched while applying for a driver's license to be and that database searchable by the fbi to match the fbi work diligently with the state representatives we did so under the state's authority to allow photos to be used for criminal investigations and also abided by the federal drivers license federal protection act and we consider that a very important process to access those photos in the federal agencies to make you say the state authority allows you to do this but one
2:55 am
question we have been asked over and over but if any elected officials have anything to do with the existence? so another words where is that authority coming from? with so many citizens whether elected officials have anything to do with it. >> and then to sign the ou. and this is state law prior to facial recognition we are just leveraging that state law it is already in place with the attorney level at the state level.
2:56 am
>> and with that facial recognition coming into existence, i am just wondering whatever lies you are referring to is like facial recognition. >> it is my understanding the states establish those laws because of fraud and abuse of drivers license reviewing each of the state laws to be sure we can leverage that because there are laws that are out there without the facial recognition and now the fbi has decided it would take advantage of those lots. is that a fair statement?
2:57 am
the federal driver license protection act and then obtained with motor vehicle record without official function. >> we have seen significance to provide this level of access it with a may 22nd hearing but that fpa one - - fbi use of the database is that correct quick. >> i'm not aware of that. >> it is accurate. so how many states have provided this level of direct access to the fbi? . >> we don't have direct access first 21 states.
2:58 am
what we did over the last two years that the federal and state authorities. >> the fbi had plans to increase the numbers and the database quick. >> that is up to the state. we have reached out but it is up to the representatives and it is optional so you have that level of access to the database with those policies we are aware those systems and any changes that are made. >> it is extremely clear to each of the states how it will be used and also photos coming back we ask that they purge.
2:59 am
>> how do you make them aware? is not expanding the information and available so can you describe these negotiations. >> i'm not aware of any current negotiations right now with the department of state are there any limits to this access and in accordance and then followed by the fbi and
3:00 am
what about any american with a passport quick. >> for an open assessment by an active investigation only by the fbi. >> i thinks the chairman to this important issue to the forefront i know we don't have border patrol here and the department of transportation use that technology so can you give us with those partners who are working with quick. >> with the state law
3:01 am
enforcement with the law enforcement into the fbi offices and some of those successes are assisting with their capture of the terrorist in boston and putting the pieces together to identify where pedophile is and also assisting on the ten most wanted list for homicide. >> so the greatest success with cbp we leverage travel and verification system for biometric verification at the checkpoint we are doing this solely on a pilot basis indicating a very high positive maturate at the
3:02 am
checkpoints so with that fax of accuracy with facial recognition can you give us some idea how that would be much more accurate in that application. >> the most recent we have conducted over previous test in 2014 to demonstrate those limitations with accuracy the most recent test results will be published this month but the results and accuracy
3:03 am
across the board. >> somebody finding in different algorithms quick. >> those accuracy rates that we see with participants who submitted algorithms where 70 participants in testing, the best algorithms are performing at a rate of approximately 99.seven in terms of accuracy. there is still a wide variety of wide very - - variance some of the participants fared significantly poorer than that but the best are between 99 and 99.seven. >> are there algorithms you have tested for law enforcement quick.
3:04 am
>> we don't make recommendations we provide the data necessary to make informed decisions about how an algorithm will perform in a field so for law enforcement the accuracy rates are one important aspect that needs to be considered with their other aspects taken into consideration for the acquisition of such. >> going back the bias can be built is that true quick. >> it is true the algorithms depend on the way they have been developed can have a bias associated with them and in many cases the improvement that we see that dramatic improvement comes from a transition the deep learning algorithm that has made the
3:05 am
difference. let me be clear we evaluate these as black boxes so my assertion is from the discussion with vendors and not examination of the algorithms themselves. and the training of those algorithms determine the level of the bias that may exist. >> thank you mister chairman thank you for hauling the second hearing on facial recognition it's good to have bipartisan interest on this issue i certainly understand the dynamic at play with an active investigation ongoing with the mugshots of known criminals but according to the
3:06 am
biometrics roadmap in september 2018 tsa is expanding the use of facial recognition technology to the general flying public. and in specific locations but the general flying public and the use of technology upon domestic flights which would capture the faces of mostly american citizens. so what is the legal basis? that we would have probable cause that where does it find that justification of that identity?
3:07 am
. >> in accordance with the faa act tsa has to positively identify passengers boarding aircraft. >> i will stop you right there. we fly a couple times a week. now you have to have a certified license and now we have much more accurate licenses surrendering that in the airport you have a ticketing issue so now that you say these pilot programs will hurt people and you say voluntarily but i can imagine like you have done with pre- check you will surrender your
3:08 am
right anonymity or give up the fourth amendment rights is that the dynamic that is going on quick. >> to the general traveling public using a one / one matching capability you stick it in machine and it identifies the credential if your image matches the image on the credential and gives a match result that allows you to proceed through the checkpoint we will always have the option to do that manually. >> but you have to have that data on board in the technology to begin with. right quick. >> that is embedded in your credential. the photograph is on your drivers license it is a digital recording of that
3:09 am
image and when your picture is captured by the camera it is matched to the photograph on the credential that is in a checkpoint for a database search that is the one / one what we use. >> you don't anticipate using a database for information with which to identify passengers quick. >> for international travelers and for tsa pre- check passengers also providing a passport photo we will match that to my gallery but for the general public. >> what is the size of the gallery? if anybody engages in international travel? or foreign nationals quick.
3:10 am
>> the gallery we use right now and has a photo on record we have a 20 million individuals with opm social security numbers, everything they submitted with federal documents and stolen by the chinese i'm just curious and i'm concerned that we don't have a great track record here with personal information. >> in the cybersecurity rules what we take very seriously. >> i hope so i yelled back i ask unanimous consent to enter
3:11 am
into the record with the security industry association cybersecurity providers in general knowledge with this emerging technology era it is important for quite some time and in 2005 as a police officer in the city that i patrol with access to a series of cameras disguised as a transformer on the electric pole. we had large numbers of violent complaints in the city and the citizens themselves wanted this investigated.
3:12 am
and this was 15 years ago. license plate readers we have those right now. i'm sure you are familiar with those we use them from sea to shining sea. if your vehicle is on public road these cameras are not available and then to be the cross-reference of the dmv. and then with those scores of thousands across the country.
3:13 am
and 11 smart cameras connected to software to high resolution digital cameras to determine if it is a familiar person or not. that the, one - - the cameras have learned it is constant with no alert sent it is not a familiar person that a human being receives a prompt to look at that purse one - - look at the camera. this is technology that exist and we all have. everyone here wants to protect fourth amendment rights and privacy rights of american citizens. none of us want the constitutional protections violated. but the fact is this emerging technology the facial
3:14 am
recognition is reflecting the advancement of our digital technology that we have already employed across the country including airports. like any technology there is a chance for abuse we feel those policies and procedures are extremely important. >> we require all state law enforcement entities to adhere to the required training. >> so that technology, these
3:15 am
cameras don't make arrest they just add to the case file or to the strength of the investigation of any human being investigated to determine if you have probable cause for arrest quick. >> we don't capture real-time a photo has to be submitted by law enforcement for a law enforcement purpose. >> as it should be the potential abuse of this technology and the point that we should clarify that as human beings are the investigative effort and that technology used with any criminal investigation would you concur quick.
3:16 am
>> we are very strict on the use of our system of those law enforcement entities. >> my time is expired. . >> we take this very seriously we went out to the advisory policy board with those tribal entities and talk to them about the gao findings against the first and fourth amendment and require state local tribal entities and then to restrict the access and then they must follow those standards and those working group standards.
3:17 am
>> thank you for conducting this hearing of the witnesses for being here. . >> that the fbi make changes to transparency with that facial recognition technology gao released a letter to the department of justice to recommend doj has a privacy impact assessment and not published as required with corrective action. doj did not agree with those recommendations and does not
3:18 am
fully implemented by gao. can you explain the importance transparency coming to the fbi use of facial recognition technology? . >> as you mention three of these are related to privacy three were related to accuracy. only one is closed to focus on that privacy impact assessment and that is a requirement under 2002 to be conducted to help determine the privacy implications so the doj has disagreed with that but disagree with the recommendation they have to
3:19 am
submit and that the warrant is required under the privacy act anytime it pertains to the system they have to make that information publicly available so the public knows what is going on. we stand behind those recommendations. >> and those documents have not been made public. >> so explain why the fbi they agree with these recommendations quick. >> doj disagrees with the assessment of the legal requirement publishing both reports with a facial recognition with privacy attorneys embedded in the process with the protocols and the updates and with the
3:20 am
updates to gao. so what steps do you take to protect privacy? remake the fbi monitors the audits of the state entities for system requirements and provide outreach and not have any violations or notice from the public they feel like their rights are violated. >> to what extents do you feel you take the steps with the public quick. >> they are on behalf of the department of justice i have to take the question back to them. >> please give us their respons response. >> if they are not fully compliant with the use of facial recognition but when the fbi arrest a lead
3:21 am
generated by facial recognition is that notified? . >> those are through the open assessments and they are done so following the attorney general guidelines. >> county times has the fbi with that. >> is part of that criminal investigation is not part of the process. >> if it gets through discovery? . >> the fbi services union that i represent through west virginia provides a candidate back to the field office and they make the determination if
3:22 am
there is a person of interest they are looking for. >> they are looking for other candidate matches of discovery? . >> i'm not aware of any other information other than a candidate from the interstate photosystem. >> what steps are the fbi taking to make sure that this technology is transparent by showing proper notification quick. >> providing policy and procedures to local entities they must follow and follow the standards we have established to make sure they do so in accordance with law enforcement purposes. >> how do they know if that is
3:23 am
subject to searches? . >> the law enforcement entity would have to have the authority to do so for criminal justice purpose in order to access the photosystem. >> my time is expired. >> did they meet all the requirements of the government? . >> as i mentioned early. >> did it meet all the requirements? yes or no. >> no. we still have open recommendations. >> to the f vi publish that assessment in a timely fashion in 2011? . >> no. >> did the fbi follow proper notice file proper notice in a timely fashion with facial
3:24 am
recognition technology? . >> no. did the fbi conduct proper testing of interstate photosystem? . >> proper in terms of accuracy for use? no. did the fbi test the accuracy of the state system? . >> no. . >> no. they did not file assessment notices like it was supposed to edit and provide timely notice it didn't conduct proper testing of the system and didn't check the accuracy of the system interfaced with. correct? those five things that did not do. >> that is correct but you say you have strict standards. you can count on us. we have memorandums of understanding with the states to safeguard people. that's what you told us.
3:25 am
but when they stand up the system there were five key things they had to follow that they didn't, my understanding is they still haven't corrected all of those. >> that is correct. >> and the things they were supposed to do. >> we have open recommendations. >> so we're supposed to believe everything is just fine. don't worry for you have even got to the fundamentals yet not even the first amendment concerns, fourth amendment just the process to implement the system you said earlier to the chairman strict policies that we follow. how do we have confidence in strict policies that you will follow when you didn't follow the rules in the first place? . . . .
3:26 am
and actually since the last hearing in 2017, the fbi went back and evaluated the current algorithm again and all the sizes, and the accuracy posted above and 90% of top percentile then we had reported initially. we do care about the accuracy and the testing. >> earlier you said that the focus was on memorandums of understanding between someone at the fbi signing a document and someone in the respective states that allow access to the database they garner those
3:27 am
authorities. >> some person designated to find a way i know like in ohio two weeks ago my guess is eight, nine, 10 million of them drive. access to the 9 million. >> the state authorities documents everyone can get access to. when you review the documents very carefully and talk about the use of the data and make sure they are in accordance with the federal drivers license privacy protection act as well. >> coming back to the basics the key things they are supposed to do to implement the system
3:28 am
dating all the way back to 2011 if i read the material perfectly that they didn't follow and get don't worry everything is just fine. earlier we learned two weeks ago in an environment where there are 60 million surveillance cameras around the country i appreciate the chairman having a second hearing on this and his willingness to work with the party to figure out where we go down the road. what is your disagreement by the way with gao? you said there's a disagreement. >> with regards to privacy, the doj disagrees with the assessment of the reporting of such, but i did have to take that specifically back to respond. >> thank you very much.
3:29 am
>> thank you mr. chairman and the ranking member and the panelists for being on this important hearing. i have read that facial recognition technology is susceptible to those that have grave ramifications for certain vulnerable populations. i've read for some reason it's more difficult to recognize women and minorities i'd like a private meeting with members that are interested on why this was reported and if it is reported correctly but what i want to do this follow up on the ranking member's question on the scope and accountability. how many searches ha as the fbi iran in the next generation photosystem to date? to april of 2019 there were
3:30 am
152,500 searches. >> does the fbi track if the results of the system is useful in the investigation? >> we ask them to provide feedback on the services that we provide to date haven't received any negative feedback is there any proof the system is helpful to law enforcement has that led to a conviction, gave it to me in writing. how many of the searches have led to arrests and convictions, do you have that information? how many of the fbi searches have led to the arrest of innocent people?
3:31 am
>> the ball enforcement entity must have their authorized access to the system and must do so -- >> has it led to the arrest of any innocent people? >> not to my knowledge. >> are you tracking the number that have led to the arrest? you don't know anything about any innocent person being arrested. >> maybe we should change the system as we need accountability on if the system is working or not or if it is just abusing evil. it contains over 600 million photos of individuals that are permanent primarily of people that have never been convicted of a crime and my question is why does the fbi need to gather photos of innocent people? >> we do not have innocent people or citizens i of the database. we have criminal mugshots
3:32 am
associated with criminal arrest. >> the information i read in the paper must be wrong. i'm going to follow-up with a letter for clarification because i told you have 600 million in the database. it is important whether we know the use of the technology leads to any benefit for society especially determining whether it is helping to solve or are we weighing in on the constitutional rights of people and creating the constitutional risks. unless there is a sufficiently database for law enforcement and so my question is what are the current reporting requirements regarding the use of facial recognition technology? is there any oversight reporting requirements on the use of this
3:33 am
technology? >> the fbi monitors appropriate uses. we have a robust audit -- >> do you have a database that tracks whether or not this is working. do you have a database that tells us that the program is doing and what the penalties are to our society. i'm going to go to work on one right now. i agree with the questioning of the minority party leadership on both that you don't have answers on how it's working and how it's set up and what's coming out of it and whether it is hurting
3:34 am
people, helping people, but you don't have that information on whether it is aiding law enforcement and their goal for hunting down terrorists. we need more accountability, and i yield back. >> i recognize the ranking member. >> from the consent consumer technology to the chairman about this issue. >> without objection, so ordered. >> thank you, mr. chairman. did you test the accuracy rating i can imagine the way the algorithm fails one would be a false positive and one would be failing to recognize an actual match. which number are you reporting?
3:35 am
>> the accuracy of 99.7 i believe. the accuracy for siblings? >> we do have perhaps some of the most relevant data that i can give you is we do know that there is an impact on twins in the database. >> let me give you the data plane but i have, one is two and a half years younger than the other and he can open his brother's phone. they don't look that much alike. they look like brothers. he throws his eyebrows and changes the shape of his mouth to the way he thinks his brother monks and over this phone every single time. so, that accuracy is not 99%,
3:36 am
that is 0%. that may be an older algorithm i'm sure they improved. i want to submit for the record an article by thomas brewer called the book into a bunch of phones with a three d. printed head. >> thank you mr. chairman. so, it's accurate for certain conditions like somebody wearing a mask or makeup or maybe a sibling the accuracy does not approach or may not approach 99% to somfor some of these algorit. what do you think? >> the situations you are describing are situations where there is intent to deceive through liveness. >> is their intent to deceive in the world? that's what we are worried about at tsa.
3:37 am
with regard to something else here and this question, does the supreme court case help the due process rights require the government to promptly disclose potential exculpatory evidence so in the case where multiple photos are returned or there may be buying possible matches, does the defense get access or knowledge that there were other possible matches? let me give you an example. in a prayer hearing i had somebody testify to us where a person with 70% confidence was the person they ended up charging even though the algorithm for somebody else was up 90% so they charged the person with the algorithms that was 70% likely and passed over the one that was 90% likely in this case. can you guarantee us the fbi
3:38 am
would provide that type of information to the defense? >> first the fbi doesn't make a launch. we provide an investigative lead to the law enforcement partners but with all evidence obtained during an investigation -- >> do you ever provide more than one we? >> sometimes, yes, sir. a two or more. some states want 20 candidates had some want just to. >> does the defense gets bondage to the access and do you assign a probability or competence bubble with that facial recognition? >> i think that the prosecution team must determine on a case-by-case basis. >> so you are not sure if they get back. >> we don't provide a true match identification that. it's up to the law enforcement entity to make that decision. >> how many photos does the database have access to
3:39 am
including the state driver's license database. >> that changes daily. >> millions, tens of millions? >> i can provide to you. >> do you have access to kentucky's database? >> i can check. yes, we do. >> you have access to all of the photographs in the database and kentucky. which elected officials agreed to that? >> i believe we worked with the state authorities in kentucky to establish. >> that's not an elected official? >> the state authorities public and its predetermined and established prior to this recognition. >> you said that the walls were passed before facial recognition became -- >> they were. >> is a problem. >> thank you very much. mr. murdoch. >> thank you mr. chairman.
3:40 am
in may of 2016, the gao made six recommendations to the fbi, three related to privacy which i believe one was implemented and related. can you talk to me about the five that were not yet implemented? >> the three related to privacy focus on developing the process so that it's more a lined with the requirement. the other focuses on a timely manner so basically developing a process to make certain that those are published in a timely fashion. the other three are accuracy related and they are about expanding or testing the candidate list because as you know, the list size we took issue with the fact they didn't test the smaller list size so
3:41 am
that's one of them. the other is regularly assessing whether it actually meets their needs so that is an accuracy concern and the other focuses on the face database so those are also meeting the needs. those relate to accuracy and speak to this conversation here. the information that the fbi is using needs to be accurate especially if they are using them for their criminal investigations. it's important that the information be accurate. >> and the recommendations are made three years ago. is the lack of implementation, why has that been the case for three years? >> that probably is the question left better to the fbi. >> you stated 99.7% plus accuracy but that is specific algorithms. when you look at the algorithms are used, i assume based on your statement there are accuracy rates much lower than that. can you elaborate on that?
3:42 am
>> the range of performance in terms of the algorithms is pretty broad. some of the participants have made substantial progress and have remarkable algorithms in terms of the 99 and above presented a false negative rat rates. others are as much as i think i believe it is about a 60 fold less accurate, but both are from a variety of sources including one university algorithm for the research participation. >> i'm going to ask you this, is there data showing facial recognition accuracy versus traditional photographs and enhanced photography?
3:43 am
>> i'm not quite sure that i understand the question. >> if it is an old-fashioned technology just using photographs versus facial recognition, is there any data we have available that shows facial recognition if a large step in the wrong direction even with the challenges we are having here? >> we also test human performance in facial recognition through comparison photographs. interestingly what we find, and i refer to my testimony if you combine the two humans you don't do much better than any individually. if you combine the algorithms you don't do much better than any individual. if you combine a human and facial recognition you do substantially better. >> to put it back as to why they
3:44 am
haven't implemented the other recommendations. >> recommendations regarding the doj disagrees with the legal assessment. we have to privacy attorneys embedded in the process the whole time we published and continue to update those accordingly and we provided updates. with regards to the candidate list size since the last hearing in 2017, the fbi conducted a test of the current accuracy in the system at all of the list of sizes and we were able to validate that a person age was higher than what we published. >> one more quick question. if a bad actor with bad intentions and skill set doesn't that circumvent entire process? >> we provide a candidate back
3:45 am
and use trained fbi examiners. the system combined with the examiner provides a better response back to the law enforcement entity. >> thank you very much. does the fbi use real-time face recognition on live video feeds or have any plans to do so in the future? >> know we dno we do not. >> having experimented with real-time face recognition? >> not to my knowledge. >> do any of the law enforcement partners utilize or plan to utilize real-time face recognition technology? >> not for criminal justice purposes. >> does the department believes that they have statutory authority to do real-time face recognition? >> not to my knowledge. >> does the department of justice believe they have
3:46 am
statutory authority if dave putd support the recognition? >> no sir. >> please name the companies lobby or communicate about face recognition products they would like to provide. >> we have the testing that we have done but those are the only agent he i these we are familiah and about what the urge to defend or start participated in a facial recognition vendor testing 2018. >> so in conjunction excuses the cameras and matching algorithm was developed. that's thbut the company they ag with right now. >> how many air passengers have participated in the face
3:47 am
recognition pilots? >> i would have to get back to you with a number on that for the record. >> you couldn't tell us how many participants are u.s. citizens? under what statutory authority do they use the recognition to talk about technology on american citizens? >> the transportation security act that requires us to identify passengers before boarding aircraft and proceed to the checkpoint. >> can you tell me what they use on domestic travel generally? >> i would say that it's the same as the security act.
3:48 am
>> do they consider the real-time face recognition technology. >> yes, sir absolutely bears signage that clearly identifies using facial recognition technology and to identify passengers we don't store any photographs on the camera. >> will the travelers be able to off out? >> they always have the opportunity not to participate in the program. >> you think that is true now and in the foreseeable future? >> yes, sir. >> add additional planes and airports and beyond when you drop the bag off to be placed within aircraft we can use the facial technology there and for tsa purposes on the location
3:49 am
checkpoint. >> that is in a one-on-one comparison. what are you comparing it to if you look at changing facial recognition is flooded and the necessarily the identification what were you working with for that? against the identification in the system. >> that contradicts the earlier testimony because what you said you were giving us checking the biometrics within the identification against a facial recognition but it sounds like you are doing a lot more than that. >> i understand i just came
3:50 am
back. i came through jfk and i didn't see any of the signs that you were talking about. so i guess what i'm saying is what statutory authority gives you the ability to do that and you keep referring to 2001. i actually am a moron the transportation committee and i can tell you we never envisioned any of this. and i'm looking at the very statute myself and how can you look and suggest the statute gives you the ability to invade the privacy of american citizens? you may answer the question. >> with respect to the pilot in atlanta it's international travelers into the purpose is to match the biometrics to the back of the backdrop. the photographic images captured
3:51 am
is transmitted to the system for matching and returns a result that there's no privacy informatiothere is no privacyina associated with it. there is no pilot going on that right now. solely in atlanta and terminal. he mac thank you mr. chairman. i want to follow-up on several of these questions. you, do they record how many are captured and if so do you know the numbers? >> i don't know the numbers. i would have to sit that for the record. >> also they use the facial recognition systems which may not restrict how the private corporations use passenger data. according to an august article from "the new york times," it has said it cannot control how the companies used the data because they are not collecting photographs on their behalf and
3:52 am
officials state they believe commercial carriers had no interest in keeping or retaining the biometric data and the airlines said they are not doing so but if they did, he said that would really be up to them. they said they intend to pursue innovative models of public-private partnerships to drive collaboration and co- investments. if tsa uses the systems to scan the faces of american citizens how can you ensure that the data of the passengers isn't stored or sold by private agents? >> i would have to referral for further system with respect to the public-private partnerships when we refer to them we are talking about partnering with industry, airlines and airports solely on the front and capture system so basically the cameras that are being utilized. >> but you talk about the coinvestment. >> in accordance with the authorities we are allowed to enter into agreements with
3:53 am
airports and airlines to procure equipment on our behalf and that would be the camera system only solely for the capture. imagining in the database it is a government system and right now using the tps system. >> have you thought about how you would ensure that the private data isn't stored or sold by airline's? >> when your photograph is captured at the checkpoint and the airline is sent off for matching and database that the system is cybersecurity accordance with applicable standards and we do not transfer any personally identifiable information between us and airlines we wouldn't have an answer to that question in the
3:54 am
way that we think about it. we have issued recommendations to the privacy and accuracy and if those recommendations were implemented, that would have to go a long way to meeting some of the needs of the public as well as the youth on this committee. >> can you clarify? >> we have those six recommendations. only one has been implemented. we believe if the remaining five are implemented, that would go a long way to answering the question with the privacy impacts associated with the biometrics collections identifications at the airport as well as security concerns
3:55 am
associated with it and it will be coming from the dhs in the near future. >> the "washington post" further stated around 25,000 passengers traveled through the pilot program each week according to the article even assuming it is used it as 99% accurate which they are not. the high volume of the passenger traffic would still mean at least hundreds of passengers or inaccurately identified each week. it is a cute tricks on these that are accuratel and accuratey the? in accordance, they were capturing the rates with respect to the numbers of americans that do not return a positive match rate i would have to submit for the record. >> what would be the most effective way to measure how accurate the facial recognition systems or when testing the american system's?
3:56 am
we are not expert in testing for systems. we test algorithms for the accuracy of matching the entire system as something a little bit outside of my purview. >> i understand the value of the technology but i think that we need to have some clear regulation and guidance part essential to collect and protect our privacy. we need for keith to ensure those were implemented. >> i appreciate all of you taking the time to testify today and appreciate your service to the nation. federal prosecutor i appreciate the commitment to law enforcement and what you are trying to do to keep the united
3:57 am
states and its citizens safe. i do think that there's been some very important issues and our privacy rates on both sides of the aisle and i appreciate you addressing those concerns. one of the lines of questions as my colleague from michigan asking a little bit about the real-time use of this technology. and i want to explore that a little bit further and maybe even asking a simple question is the united states government in any way based on the knowledge of anybody at the table using facial recognition technology on american citizens without their knowledge today and if so, where and how? >> the systems are not designed for real-time capture. >> to your knowledge, the united states government from your basic knowledge isn't using the
3:58 am
facial recognition technology it is using it and processing it without their knowledge. >> i can speak on behalf of the fbi and we require it for the criminal purpose only in accordance with law enforcement. >> we are not in the work that we do beyond the scope. >> it's also outside of their scope. >> do you know of any plans to use that technology without the consent of the american citizen? >> not with respect to tsa. >> they will not develop technology outside of the criminal purpose.
3:59 am
you said in response in one of the questions about the real-time use. can you expand on a? >> we only collect a photograph in conjunction with criminal justice. our law enforcement partners, the state and local, federal entities must be authorized to have access in order to search the systems. i think my colleague for yielding a bit of his time here. according to the fbi records 10,554,985 criminal arrests were made committee ran about 59% conviction rate. i think that they are witnessing he must be reminded every
4:00 am
american that has arrested by probable cause is a tool in the toolbox to add to the strength or the weakness. >> they have the option to submit a photograph in accordance with the criminal investigation. >> moving quickly, one of my colleagues mentioned there was a 70% match on a subject, and that is a subject that was arrested and those are the 90% match that was not arrested. it is not arrested mean not investigative? spec we provide candidates back. >> in the course of a regular
4:01 am
investigation is reasonable suspicion ground for investigation? spinet i'm not a law enforcement officer. >> well, iem, and it is. probable cause is the standard for a rest beyond a reasonable doubt or shadow of a doubt as a standard for conviction and i very much appreciate everyone's testimony today. this is an emerging technology. mr. chairman, mr. ranking member we should watch the technology closely and protected for the rights of the american citizens we should also recognize this can be a very valuable tool for law enforcement and to fight crime in the country. >> thank you mr. chairman. we recognize i think the advancement of science is making
4:02 am
particularly in facial recognition advancement that's what we are trying to ascertain here and yet it's being used as if it were. the fbi uses this facial recognition system and we rely heavily on the gao of course. they said the officials stated that there is value in searching all available external data bases regardless the level of accuracy. that's where my question goes regardless of their level of
4:03 am
accuracy. the fbi has said that the facial recognition tool is used for investigative leads only. what's the value of searching inaccurate databases? i can see the downside in mistaken identity, misidentification? why is there any value in searching whatever database appears to be the case is available to you based on the
4:04 am
investigative leads only. >> the fbi uses our trained examiners to look at the candidates that come back on the search for the fbi opened an investigation and it evaluates all the candidates and provides the search. >> can it lead to conviction? >> the field agent is the one that's primary to that case. they know all the details. we wouldn't be making that decision it would be up to them to use it as a tool. >> not only could it lead to a conviction that may be an inaccurate conviction. >> we hope not. >> or perhaps they would be inaccurate since we are using
4:05 am
the database or investigative purposes alone cannot alone, but as well. here's what bothers me most. there's been a study done which included an fbi expert and found the face debate could facial recognition systems one like that's sold by microsoft and ibm were more inaccurate when used on darker skinned individuals, women and people age 18 to 30 when compared with white men, so we do have some indication when we look at what the population is. do you agree with the findings of this study? >> there are demographic effec
4:06 am
effects. this is very time-dependent it depends time which the evaluation was done. we are prepared to release demographic information. >> my concern is there is excessive some would say over policing in the communities. i understand why, but it has resulted in african-americans being incarcerated at four times the rate of white americans and african-americans are over represented in mugshots and some systems scan for potential matches. do you agree that both the present overrepresentation and
4:07 am
deliver accuracy grades that facial recognition systems have when assessing darker skinned people such as african-americans that it is possible that false convictions could result in the fbi's use of these external systems if they are not audited tax >> you may answer the question. >> the fbi retains photos but they are associated in criminal arrests and a ten print fingerprint. the process with the state federal, local and tribal agencies we share the auditors out of those agencies and with the security requirements in
4:08 am
accordance with the securities and we look at the policies and procedures and standards to ensure that they are required training and following the process. >> thank you mr. chairman. i think that we are all very much aware of the effects of the survey. people's behavior certainly changes. noncriminal speech and behavior alters the way people behave just even as a pastor for many years i know how that has had a chilling effect on speech so this is an extremely serious thing where we know the possibility of surveillance is out there. has the fbi ever, you mentioned a while ago that facial services
4:09 am
universe and into that affect. it doebecause that particular ur any other in the fbi farm for images, photographs or any other type of information on american citizens and social media, whatever? has the fbi ever purchased from a third party contract or wherever else images, photographs information? >> they maintain only criminal mugshots photos. >> i would like to ask to be submitted in the record and article by the toolbox by the fbi to monitor social media.
4:10 am
i've also like to submit for the record and archive copy of the web domaiweb domain that statese software is used for the automated collection of social media data to be submitted. >> without objection, so ordered. >> i would also like to submit the software service agreement and individual user license purchased by allied associates international. i don't know where you are coming from to not be aware of that. >> i would have to find out from the other entities within the fbi. i represent the technology that is used for the criminal justice
4:11 am
purpose. >> so there's there is a wholer avenue of the facial recognition taking place but you know nothing about. >> not that i'm aware of. >> if you don't do anything to this, then there is. >> to collect information on the u.s. citizens? >> i'm only aware of the use of our time for criminal justice purposes. >> and your system would include the system of the driver's license database of multiple states. >> our system does not retain drivers license photos. >> so there's two different systems you have your internal system and the system you can access. >> we do not have direct access.
4:12 am
>> the study by the center on privacy and technology found but you do have access to the outcome of the total of 217 million which comes to about one out of every two adults but you have access to. >> that is incorrect. we disagree with that. the fbi threw an active investigation cans at a photo. >> how many do you have access to? >> weekends at a photo and that they provide a candidate back. we do not have access to those photos. >> there is a pre- crime database if you will have the fbi or any other federal agency used a real-time surveillance for some location, is that ever
4:13 am
done? >> not to my knowledge. >> the numbers, what number of photos do they have access to in just their database? it's a little over 36 million. >> been in the database that they can send information to our screamed and accused with interaction at the state level. excess across all, about 640 million. >> 640 million photos.
4:14 am
the access to over 600 some million photos that didn't comply with five things they were supposed to comply with when they set up the system and they are still not in compliance with. it's over 640 million photos. the fbi really only searches for criminals. they are looking for the criminal photos. not necessarily convicted. >> arrested by searching the databases. we would have to go back and do a survey. every 90 days we go out to the agency to see if there's any input they could provide to us.
4:15 am
we do know there are arrests made but it's not on the identification of a photograph. it's a tool to be part of the case that they have. if i could add one more thing about a 640 million. most of those are simple photos. those are primarily simple photos. thank you mr. chairman for holding the hearing. without the use of facial recognition increasing, it's important that the technology isn't rushed to market and that all communities are treated equally and fairly. in your testimony you mentioned a report that is due for publication this fall on the demographic effects in mugshots. can you talk a little bit about the report in your objective? >> the objective is to ensure
4:16 am
complete transparency with regard to the performance of the algorithms that we evaluate and to see if we can use vigorous statistical analysis to demonstrate the presence or absence of the democratic effects. that'that statistical analysis t been completed yet in that preliminary data that have suggested that the demographic effects of a difference in sex and race can affect and have differences with the algorithms we expect diminishing the effect
4:17 am
overall. so come into full we will have the final report of the demographic effect. >> when you are doing a calculation for the companies argue testing for a demographic consistency? >> we don't test for specific companies on their behalf. we evaluate the algorithms that submit to us through this voluntary program. so probably don't test specifically for algorithms demographic effects. we are talking about the demographic effects across all of the algorithms and they are submitted. >> what are you doing to make sure no categories of people are suffering from the lower rate accuracy.
4:18 am
for the data about the level of the demographic effect we have no regulatory authority to do anything about that other than make that data available for the policymakers to make. >> do you have a comment about that? tsa has been partnering with the biometric for international travelers. how much training that the operators received prior to beginning the program at jfk and lax? >> the training was significant. i would say multiple days of training and how the system worked and how to analyze the results and effectively use the system. >> what were the top complaints that we see during this pilot? >> i am not aware of any specific category complaints that rose to the surface by
4:19 am
using the biometrics. >> any complaints by employees? >> i would say employees in general when you introduce a new technology that change can be somewhat challenging of having been down to atlanta and talk to the operators down there as well as the director in charge of the airport, they embrace the technology to be a significant enhancement to security at the checkpoint the report has been compiled, yes ma'am. >> thank you very much. i yield back.
4:20 am
>> i'm not going to beat up on you but i want to come back and give you two pieces of advice. one is the same advice i give to every witness that sits in the seat right next to the gao. if they didn't have the i'm not happy so here's what i would recommend on the five outstanding days with the five recommendations they have, are you willing to do that? the fact you only close one of them out prior to the last hearing is what i understand. if that's not accurate? i can tell that you were smiling so do you not agree? >> we have been completing audits into the completed 14 of the 21 and i think that they felt that was enough to satisfy. >> if you will report back to the committee the progress we
4:21 am
are making when it comes to that we want you to have all of the tools to do what you need to do. one, the second thing i would mention you mentioned not having any real-time systems and yet, we have testimony a couple of weeks ago from georgetown but indicated the chicago police department, detroit police department has real-time. they purchased it where they are actually taking real-time images. do they ping the fbi to validate what they picked up in real-time and with what you have on your database? >> there are authorized entities that have access to the system greatly trained them and expect them to follow the policies. we have police departments in chicago and detroit and then getting you to authenticate that through the databases that
4:22 am
correct? >> days at a photo -- >> but with real-time surveillance. that is opposite of the testimony. looking at actual real-time facial recognition when the travelers come in and out, so i guess you are saying that right now you are not doing that in dallas anymore. i can't comment on the program because they do it for the end her and exit in international travel. after all of the priorities that the tsa has and all of the inefficient that this committee and other committees have,
4:23 am
facial recognition certainly cannot be the top priority in terms of what you're looking at to make sure the traveling public is safe. is that the top priority in terms of the achilles' heel? that isn't a question tha the qi asked. is that the top priority? >> multiple priority is for tsa. >> what is your top priority? there can only be one top. this is a softball question. >> for the enhanced properties grinning at the checkpoint for the better assessment of the kerry on baggage. >> you have taken photos of american citizen dropping off their bags. is that correct? you talked about the fact it's not just a one-on-one and we are taking photographs of people
4:24 am
that have backdrops. is that correct? >> will maybe choose to participate and one. >> i've flown out of that on delta so you can guarantee that i was not photographed as i've never given anybody my permission on international travel to my knowledge he can you guarantee that i have not picked up on that? >> unless you were photographed dropping off the bag. >> i gave no one permission to take my picture dropping off my bag. i am an american citizen. what right do you have to take that photo? >> you shouldn't have been photographed. >> and you can't guarantee that i wasn't. here's what i would recommend it. i am all about making sure that we have screaming, but i can promise you i have gone through
4:25 am
screenings more than most americans and there are inefficiencies in the problems that have nothing to do with facial recognition and a tow you get that right i would suggest you put this pilot program on hold because i don't know of any appropriation that specifically allowed you to have this pilot program. are you aware of any? you keep referring back to a 2001 wall and i'm not aware of any appropriations that have given you the right to do this program. >> i'm not aware of any specific. >> i would recommend that you stop it until you find out your statutory authority. and i yield back. .. will happen so often is that
4:26 am
people say they will get things done and they never get done. so mr. meadows, in the spirit of efficiency and effectiveness -- every reasonable request. that they get together so we can get some of these items resolved. i'm going to call you all back in about two months maybe, i'll figure it out. because i'm worried that this is going to go on and on and in the meantime i'm sure will be able to come up with bipartisan solutions. but the american citizens are being placed in jeopardy as a
4:27 am
result of a system that is not ready. i will call you all back. i hope you all get together as soon as possible again i say this because i see over and over again and will be in the same position or worse in three years, five years, ten years. and by that time so many citizens may have been subjected to something they should not be. with that, i call on --'s. >> mr. chairman i appreciate your leadership and follow-up. >> no problem. and now: distinguished lady from michigan, ms. lawrence. >> thank you, mr. chair, doctor iranian, do you think the third-party testing is important for the safe deployment of facial recognition technology? i want you to know that i did on
4:28 am
the committal justice appropriation committee and funding for nist is something i have a responsibility. i would really like to respond to these questions. >> i think independent assessment of new technology particularly if they're going to be used in certain ways is an essential part in one of the things we are privileged to do. >> how dependent are government agencies on nist finding? how dependent? >> it's hard for me too assess that. i think we certainly have collaborative relationships with dhs, fdi, with other federal agencies and part of our statutory requirement is working with other agencies on
4:29 am
advancement of technologies and evaluation of technology. >> is a way that we can move forward that you can do an assessment so we would know when were talking about the findings which is a critical factor right now. is there a way we can move forward so we can assess what is the role that is played by the third-party? >> with respect to facial recognition, we have ongoing evaluations on a rolling basis. participants can submit algorithms at any time. we continue to provide open public transparent evaluation methodology so everyone, federal agencies and the public, private sector can see the results of our testing and make a determination on effectiveness
4:30 am
of the algorithms. >> through the chair i would like to see and review those. which organizations are currently equipped to accurately test new facial recognition technology? >> we are certainly equipped to do that. at mist, i don't have any information about any other entities that might be equipped to do that. >> do you believe that nist currently has significant funding and resources to carry out in state or barrier official recognition industry? >> yes, we have sufficient resources today to be able to execute the program that we have in biometrics. >> to carry out that you are saying this is an evolving and
4:31 am
were looking at the challenges you have enough funding for the rnd and for the checks and balances for you to be the standard facial recognition industry? nothing frustrates me more for you to come before congress and say have everything i need and then when you don't do the job well we didn't have the funding. so i'm asking this question and i need you to be very honest. >> i'll make two remarks, one is, we have a long track record of delivering high-quality evaluations in biometrics for nearly 60 years. second part, it is a bit awkward for me in front of congress or any federal official to speak
4:32 am
about funding ovals, i will make the comet that any resource organization can do more with more. i'll leave it at that. >> will for me too do my job i have to get accurate and you have to have a plan and direct us -- >> is anyone on the panel wanted to comment on the organization and the ability to accurately test new facial recognition technology? are there any comments no. thank you. >> thank you very much. mr. miller. >> thank you, chairman cummings and ranking member jordan and thank you all for being here today. america has been a leader and innovator in the technology sector. american companies have pioneered any of the technologies deployed around the
4:33 am
world. however, as the sector continues to grow we need to ensure our government agencies are accurately deploying technology within the bounds of law. this past week, i was in china and i saw facial recognition technology deployed on a massive scale. from the moment i was getting 80 to get on the airplane. there were cameras everywhere. allie bubba recently instituted a program where customers can smile to pay. using facial recognition technology. i also saw cameras at street crossings that can camp pinpoint certain individuals who are breaking traffic laws, it was daunting to see the government shaming individuals so publicly. which is a stark contrast to what her privacy and liberty is in america. they would flush your face are here. the facial recognition technology poses many questions on the state.
4:34 am
about the appropriate use of this technology. doctor, what steps can a government take to ensure emission technology is being deployed in a way that it is accurate? >> thank you for that question. i will always go back to the recommendation made when we did this work a few years ago at d.o.j. is still accuracy, transference he are key and vital to when were talking about technology. as well as we are protecting privacy rights. to go back to the recommendation we want d.o.j. to pay more attention to the testing and regularly assess whether the ndi, it f whether the information is accurate and we want them to assess and have some understanding of whatever information that they're getting from external partners are also accurate.
4:35 am
>> thank you. to your knowledge has fbi had any misidentification of individuals when utilizing facial recognition technology? i would like to go back, we did test in 2017 the fbi did test they saw improvement in accuracy. they had facial recognition with test and implementing a new algorithm and we were continuously work with state and federal local partners on the use of the system and we also commissioned a 2019 and onward it's called ongoing facial recognition.
4:36 am
his ministry to buy the next generation and edification. so the system was divined to return to her more candidates we provide an investigative lead back to law enforcement the law enforcement entity we require training by law enforcement to follow the ndi interstate policy and implementation in the facial identification scientific working standards. anyone running a search through the ndi interstate photosystem must provide what the policy and standard and audited by fdi. >> can you discuss what's in place to allow the facial technology and how strictly they are enforced? >> i know for the fdi space service unit, and fdi office must have an open and investment
4:37 am
and follow the guidelines associated with that for us to be able to receive a photo from them and a photo for search. >> okay. doctor, to your knowledge has fdi been adhering to the regulations? >> we are working closely with the fbi if i could go back to something she said earlier. the testing that they're currently doing, then information they're providing until we see that we will not be closing recommendations. we need to make certain that their meeting the recommendations as we put forward. >> thank you. i yield back my time. >> mr. gomez. >> thank you, mr. chairman. in the history of this country we have always had a debate and a goal of trying to balance
4:38 am
security with liberty. but in the air of facial recognition i feel worse doubling into the future without really understanding how much liberty we are giving up for how much security. it is really with the understanding we have to set up guidelines that really dictate the use of this technology. so that is where my approach comes from. i have a lot of concerns regarding the positive rate of technology, racial bias in the technology, gender bias and even during, pride month, june is pride month, i think about the past gender and 99 dairy communities, we've seen reports that show that black women are more likely to be misidentified than any other group. when you layer on top of that the trans gender non- binary black individual what happens to
4:39 am
those results? >> have you seen any data when it comes to the lgbtq immunity exists specifically the transgender community? >> we have not done an analysis on accuracy rates for trans gender communities, i'm not sure how we would obtain the relevant data that we would used to do that. i am aware, i've been made aware of concerns of the transgender community about the potential for problematic use here. >> i appreciate that. a lot of this is revolved around training. in another of indicated that some people are likely to be believed pewter generated results. those who are not specially trained in face recognition have
4:40 am
prominent i' i did find people y don't know even if they perform face identification as part of their work. i am kinda keeping that in mind with my question i'm about to enter. first, what is a constant interval level of the fbi that uses when it comes to running the program for the matches? is 80%, 85%, 95%? >> accorded accuracy, and we do not have matches, let me clarify that. it's an investigative lead, two or more candidates, our system is not built to respond to one response. currently we have an 85% accuracy rate although since the last hearing. >> that's not what i'm asking, when you run the program is it set to a high level that it needs to be accurate to 95% level that the computer recognizes that this individual is under 5% likely to be this
4:41 am
person wears 80%? like amazon, they sell the program at 80% default, what you run your program and? >> we don't conduct a nano defecation match, we don't look at that. we have an accuracy rate that we rely on and we are currently implementing the new vendor recognition test result at 99 but when 2% and it's 99.72 uttering 50. that is the new algorithm. because is not a true identification -- >> how does human tendency trust venerated results? >> well through the testing for sure. we also use other agencies in entity universities to provide testing results to us. >> you trained the fbi personnel to perform facial comparisons of persons that are unknown to them? >> we received pro photos from active investigations from the
4:42 am
fbi field office, fbi agent and they process that photo against the mug shot report and received a candidate back and they are trained to evaluate. >> so is fbi trained personnel in accuracy official recognition algorithm? >> bias? no, sir back the fbi does publish, winds at? >> i think the employees, our system does not look at skin tone and futures preacher under. as a mathematical computation that comes back. there to look at the mathematical phase. >> i understand that you're basically describing facial recognition technology, but outside studies have shown there's a bias when it comes to concern population. in the air rate is a lot higher. and the aclu conducted a match
4:43 am
of different members of congress that 80% confidence and members of congress including myself were mismatched positively -- >> the technology you're referencing to is a match and we did not do that. >> seated broader? >> we do to stew 50 candidates or more with a one-to-one match, it's not a match. >> in a manner consistent with the guidelines and recognitions but official notification working group. the official identification group does not endorse the body in person to compare the a.i. certification the since there is no formal certification process that the working group endorses, what standard of fdi official
4:44 am
and also. >> have to comply with as well and we require all entities that have access to the interstate photosystem to follow the policy and implementation guide in the standard. if default does. >> your time is expired. >> thank you mr. chairman. >> mrs. presley. >> thank you, mr. chairman. it is been abundantly clear that facial recognition technology fund by design unlawfully producing false matches due to algorithmic science. including everyday americans and in fact congress which represented gomez was one of those. which is speaking to that. there is growing incredible concern over the unauthorized use of the technology in public spaces such as airport schools and courthouses. these systems could certainly be
4:45 am
subject to misuse and abuse by law enforcement. we know this technology is often used without consent. in that there are, no real safeguards, there is no guard was here. this is not fully developed, i want to take a moment to say i appreciate the leadership of the city of somerville in my district of massachusetts seven who have passed the moratorium on the surveillance and on the software because of the fact that it is not developed and another safeguards in no gardens. much of my questioning is already been asked but i want to pick up on a couple of things in the space of consent and i wanted to give them accuracy questions and better understand the purposes of the record here. you keep data on how many people
4:46 am
opt out official recognition technology? >> no i'm not aware we are collecting data on people who choose not to participate. i do not think were collecting it no ma'am. >> okay. you're annoyed you have any people had opted out a previous tsa, facial recognition pilot programs pick - no ma'am. >> you know how many passengers were notified of tsa use of facial recognition technology? >> the notification at the airport of the signage and verbal instructions from the officers so if they're in elaine being piloted i would say one 100% of the people being aware that is being used. and you have to assume a suitable post to actually have the camera capture. >> again, if this is based on signing which are many ways can
4:47 am
be arbitrary. how are folks even aware of the options to opt out. >> for your identification, and for the identification. >> is not communicated in multiple linkages? >> for the purpose of pilot has not been communicated. >> again for the purposes of the record, i over spoke based on my own desires that the municipality passed an ordinance to band that is not past the order trained so i wanted to correct that for the purposes of the record. let me for moment get back into regarding for facial
4:48 am
recognition. are you aware of how many government agencies use or possess fickle interfacial recognition technology. >> anyone? >> i don't other. >> no doonor do i. the gml does have more group work right now looking at the use of authority and tsa. we will be following up on information here. so is there comparative benchmark to the accuracy of these programs and how they compare with other programs? we are not aware of that as of yet. >> okay. did nist present any red flags to any agencies in any particular system used by government agencies that you're aware?
4:49 am
>> this does not interpret the scientific data in terms of red sox. we just ensure that everyone who is using facial technology has access to the data that we published openly about the performance of algorithms that have been voluntarily submitted voluntarily. >> i think that's it for now ideal. >> let me ask you this, this ongoing work, what is happening there. >> we have ongoing work at the request of both the senate and house homeland committee to look at the use of face recognition technology but you just in particular tsa and cbp. we also have ongoing work looking at the commercial uses a face recognition technology and if i circle back to congresswoman presley, about
4:50 am
consent, there is a senate bill that will look at consent that only looks at consent from commercial usage not by federal usage. >> we have those ongoing jobs and then gal does have a request to look at face recognition technology across the rest of law enforcement - going back, questions about the whole idea of language, do you feel comfortable, i'm assuming the elected tsa already. >> we are starting the engagement so we -- you not look to the private program. >> not as of yet but i imagine that will be part of what we examine. but that engagement just started a gal in one of the things i'm hoping that you look at is the whole question, people are trying to get to where they
4:51 am
gotta go. a lot of people don't even know what facial recognition is. then if you have a language problem. that is even more and something to consider. have you thought about that quick. >> yes, sir, i was going to answer questions, one of the reasons during the pilots is to efficiency for how we communicate with passengers per can we do better to the signage be better, multiple language is something we should be looking at. all of that will be assessed with respect to the pilots before we make a decision moving forward. >> mrs. lee. >> thank you mr. chairman. i hope this is okay, this stuff freaks me out, i'm a little freaked out by facial recognition mr. chairman. i hope that's okay. >> that's okay. >> thank you. my residence in michigan and
4:52 am
congressional district increased surveillance and overpricing for decades. currently the city of detroit reeled out real-time video surveillance program called the project greenlight in 2016 to monitor crime a late night businesses like gas stations and liquor stores. but now the system has expanded to over 500 locations including parks, churches, schools, women's clinic, addiction treatment centers and now public housing building. without notice or public, from residents the detroit police department added facial recognition technology to the project greenlight which means detroit police apartment has the ability to locate anyone who has a michigan driver's license or an arrest record in real-time using video cameras mounted across the city and the database of over 50 million photos. in january 2019 reports emerged the fbi had begun up highlighting use of immunize amn
4:53 am
recognition in the amazon controversial software that can match faces in real-time video. some like to project greenlight, recognition like surveillance programs has dangerous high error rates are women of color compared to white males. in the 13 congressional district, residents will be dissed personally bear the harms of face recognition identification. so what policy does fdi have in place regarding the use of real-time facial recognition technology, and her claims that you all are not using, but there's a pilot program correct. >> there is not, for the amazon recognition software, to the best of my knowledge and i verified before it came today, the fdi doesn't have a contract with amazon for the recognition software. we do not perform real-time surveillance. >> if you could produce the documentation in the information to a committee i would really greatly appreciate that. >> we will do so.
4:54 am
>> can explain how is not currently using amazon recognition adult? >> we are not. >> march 2017 released a report of accuracy of facial recognition to apply to individuals captured in real-time video footage. it found significantly higher air rates for real-time use recognition with accuracy rates as low as 60%. so do you think the use of real-time facial recognition technology is ready for law enforcement uses? >> that the judgment that were not prepared to make, that the policy judgment that should be on the best available scientific data which is our position. >> what is your scientific database? >> it verifies that the facial recognition accuracy is highly dependent on image quality and on the presence of injury in both of those things can affect the ability --
4:55 am
>> is a valuable solution improving the look real-time capability? >> i cannot predict how accurate the systems will be in the future as they continue to develop, currently systems that use facial images that are not in profile or not straight on or facial images that are indistinct or blurred have a much lower ability to match. >> do you have any information about the inaccuracy and i know you all have several recommendations, can you talk a little more about questions in regard, is a fixable? >> in regard to question about the amazon recognition technology, that was not something that we looked at for the purpose of our report so i will not be able to speak to that. >> but in regards to the usage of facial recognition accuracy. you had six recommendations but
4:56 am
i was talking to some of my colleagues, how do you fix something like this where you don't know innocent people into database. the numbers are 411 million and i heard 600 million people there now in this database that is being used for criminal justice processes and i'm not sure what is the definition of that. >> soul start at the beginning. for the mgi, ips there are 36 million photos in the criminal part of that. there 21 million photos for the civil part of that. i do look across all of the suitable database and retargeted. , that is over 600 million. that is what i'm talking about earlier. the recommendations that we made, the three that we made remains accuracy, we feel like this would go a long way in
4:57 am
helping d.o.j., better ensure that the data they are collecting the way they are using the information is accurate, d.o.j. has yet to close recommendations and we will work closely to get this close because the issues around that are important in the vitally important when using this technology. >> mr. chairman, this is very important to my district and the other. if we can get follow-up confirmation then indeed the current administration does not have any pilot program going on with amazon recognition program. >> thank you very much. i don't know if you heard earlier, we will bring folks back in six weeks to team up in the area and will be before then they will have those results. but definitely we will check back then.
4:58 am
all right. >> ms. ocasio-cortez. >> figure mr. trippi in the fourth amendment our founding fathers and dolls with the right of people to be secure in their persons, houses, papers and effects against unreasonable searches and seizures. the fourth amendment guarantees us that these areas shall not be unreasonably assured upon with a warrant. over the last few weeks we've been hearing from the private sector or the public that we've heard about facial recognition technology being used in airport, protest, being purchased off of social media et cetera. you're with the fbi, does fdi ever attain words before facial recognition technology? >> the criminal mug shot are searched by the law enforcement
4:59 am
partners and are collected to an arrest. in the criminal handprint and fingerprint and the use of facial recognition, it's beyond the search of the criminal database for standing a person's face is a searching their face in order to match the database does fdi ever obtain a search for that. >> we do not do real-time searching. >> to require your external partners to obtain a warrant. >> they must do so with the criminal law enforcement interest. >> does fdi use any information from any other agency? >> in with respect to facial recognition. >> we share with regard to law enforcement purposes. >> on may 22 during cummings statement his president of the baltimore protest following the death of freddie gray. as a protest the baltimore
5:00 am
police department use facial recognition technology to identify interest certain citizens present at the protest exercising the first amendment rights. has fdi ever use facial recognition deployed near a courthouse or any other sensitive location? >> no we have not. >> and do you think that the generalized facial surveillance should be permissible if you think that undermines the first amendment? >> i think protecting the american people is extremely important in the fdi absolutely wants the most fair system, we want to make sure we are following the guidelines and protocols and standards that we put in place for law enforcement. >> thinking. you are with the tsa, its outline proposals to collaborate with private companies including
5:01 am
jetblue. to develop and implement the facial recognition search system. is this correct. >> we have issued a security program amendment details identification as a backdrop. in terms of partnering with them to develop the back and matching system that is something were engaged with the entrance ebp. >> that is folks check in and get more depressed from? >> i would use the term kiosk for that. >> there is no equity there, the only way to verify that with also. we have equities on checkpoint and also at the backdrop required to ensure they match the back. >> to individuals no that is happening and do they provide explicit consent? is often. >> we bought the opportunity to not participate. >> it's often been out - yes, ma'am. >> is possible that delta are
5:02 am
working with the tsa to capture photos of passenger faces without the explicit opt in consent. >> i was in atlanta last week and watch the delta check in process of backdrop process and it was very clear while he was on the to passengers recorded opportunity if you'd like to use facial capture for identification lee stand up for the camera and will do so. there is no automatic capture of passengers. >> this capture is not saved in any way correct - no ma'am. the camera captures the images encrypted and sent to the matching system which is what cbp uses for the purpose of match and not match result is sent back to the operator. >> is that captured image destroyed. >> is not retained at all. >> so it's not retained. >> is not retained on the came camera. >> could these companies and potentially be using any part of
5:03 am
this process to capture the algorithm or data. >> no ma'am. i do not see that happening currently with the policies right now. >> thank you very much i you back. >> a much. >> thank you very much, mr. chairman. >> when we had our hearing on may 22 committee there was a researcher who was testifying about data sets and masseuses. and they may not adequately test the full range of diversity that is present in the u.s. population. she said quote, and evaluating venture mark data set from organizations that have missed i found surprising imbalances, one prominent miss data is 75% male and 80% lighter skin what i like to call appeal male data set"
5:04 am
said doctor remedy, can you discuss how representative data sets are when it comes to race, gender and age. >> sure, the data we obtain is from multiple sources, the largest amount of data that we get, first i need to make a distinction between data that we are releasing as part of the ability for vendors to determine whether they're able to submit the algorithms for system and evaluation process. too that we provide them with data for that. >> the rest of our data, the vast majority is sequestered and not made public, is solely for the purposes of evaluation. most of the data is fbi image data that we sequester and protect from release. there are some other image data related to create commons to
5:05 am
images that we have received with full institutional review that involves permissions and also deceased data set. in all cases if you look at the full sweep of data, it is true that it is not representative of the population of a whole. however, we have a large enough data set that are evaluation capabilities can be statistically analyzed to determine demographic effects of race, age, or. and were in the process of doing that now and will really set in the fall. >> so i gather since the last hearing you been testing for differential error rates on facial recognition system between races and genders. can you talk a little bit more about the error rates of the
5:06 am
algorithms that you tested between different races and genders? >> sure. i say a little bit of preliminary information but i want to stress the full statistical analysis and rigorous analysis is not completed yet. the report will be released in the fall knowledgeable conclusions that we have with regard to effects broadly speaking. we can say that there are still remaining differences even with the extraordinary advances in the algorithms over the last five years. there are still differences remaining that we can detect. we don't yet know whether those differences, whether it's with race, or age, they are significant. we don't know yet until we completed the analysis. >> so you understand the concern, there's two levels of analysis that we are somatic
5:07 am
here today. one is that the actual question of whether we like or don't like this technology given the general threat that it can pose to civil liberties. the second theme is whether recognizing that the technology is barely ahead anyhow and is being adopted and applied increasingly across many different platforms, let's say and uses. whether it's being developed in a way that ensures that when it's used, is not being used in a discriminatory fashion, is not being applied unfairly, et cetera. and that o of the algorithms beg developed in a way that is respectful of accurate data. and were not there yet.
5:08 am
so were going to be paying a lot of attention, i'm glad the treatment going to have you all come back. i think he's right that this is sort of a moving target here. were going to be paying a lot of attention to how the data gets digested and how the algorithms that flow from the data are being applied whether the accurate and so forth. so we appreciate your testimony, but obviously this is not the end of the inquiry. without a deal back. >> a while ago we were told that the basis for a lot of these agreements between the fbi and the states were -- the authorization was actually,
5:09 am
regulations, whatever were put together before technology became about. were talking about the moving target. it was not anticipating this. and we still haven't caught up. that's a part of the problem. thank you very much. mr. jordan. >> thank you. thank you, mr. chairman i want to think of witnesses for being here today. i purchased the time and expertise that you bring and brought to this important hearing. i think you understand from both sides of the high authors a concern and i appreciate being here. i know yet it's a lot of questions but i hope you understand how serious everyone is on this committee with this issue. i think it understand the framework. he talked about strict standards in place. there were strict standards in place and people from our style of the aisle. and how they go to the fisa court and get information and put information in front of the fisa court. the attorney general of the
5:10 am
united states has tapped u.s. attorney john durham ... find them from the fbi on one presidential campaign. so this is the context in the framework that many on our side see this happening and it's happening when gal, not jim jordan, not republicans, they said when you guys started this, started using this, you didn't follow the e-commerce law, you didn't do privacy impact assessment like you're supposed to, he didn't conduct private trip testing. so that's the backdrop, that's agreement. so when republicans talk about were concerned and with democrats, i appreciate the chairman focus on two areas and our third hearing. looking at legislation that we may attempt to master. this is the framework. i hope you will tell the folks at the fbi, we appreciate the debt great work in a particular country and stopping bad things from happening and fighting
5:11 am
about people who did bad things. but the framework in the context is very serious and that's why we come with the intensity both two weeks ago antedating. today, thank you for your leadership in our thinker witnesses again for being here. >> i too want to think the witnesses are being here for almost three hours. we really do appreciate your testimony. of all the issues that we've been dealing with, this is probably will be received the most intense scrutiny of the mall. the ranking member referred back to that. we also said that two subcommittees are also looking into this because we want to get it right it's not important. so i think you print without
5:12 am
objection the following should be a part of the hearing records. the face recognition performance, role of demographic information, scientific study, data december 6, 2012. , face-off, law enforcement use of face recognition technology, the electronic frontier foundation jail party commissions. department of justice letter to ag and gal, opening ongoing face recognition and vendor tests, one verification in this report in this. ongoing face recognition under test brought to in this report. phase and video evaluation face recognition on noncooperative subjects in this report.
5:13 am
letters calling for were torn on face recognition correlation letter and correlation of privacy civil liberties and for rights. thank you for aclu, georgetown law, lgbtq and the initiation team. i want to thank again are witnesses for being here today and without objection all members will have five legislative days within which to submit additional russians for the witnesses to the chair which will be forwarded for the response i want to ask our witnesses to please respond as possibly as soon as possible. with that the hearing is adjourned. [inaudible conversations]
5:14 am
[inaudible conversations]
5:15 am
5:16 am
objection. ms. collins: madam president, as the snoor senior -- as the senior republican woman in the senate today, i am pleased to begin a series of speeches, along with my good friend from california, the senior democratic woman senator, senator dianne feinstein to commemorate a significant milestone in our nation's history, 100 years ago

12 Views

info Stream Only

Uploaded by TV Archive on