Wait, What?! Toronto Police Using Facial Recognition Technology



Wait, What?! Toronto Police Using Facial Recognition Technology

Written by Stuart O’Connell (All rights reserved to author).


Facial recognition technology represents one of the most significant modern threats to personal privacy, given—in part—the ubiquity of cameras in public and private spaces, the development of the technology through deep learning non-linear data processing systems, and the nature of the information it yields: identity, primarily, but often aggregated with information detailing location and time.

While facial recognition (FR) has been in development since the 1970s, its use by government has become suddenly more viable and thus increased.

It is important as the technology gains wider use that we as a society are cognizant of the risks of the technology as well as its social value so that we can demark acceptable limits for its use.  This cannot meaningfully occur if citizens are generally unaware that the technology is being used.

Case in point: prior to the Toronto Star newspaper breaking the story in May of 2019, few Torontonians would have been aware that for over a year the Toronto Police Service has been using FR in some criminal investigations. While the use of the technology by Toronto police wasn’t secret, it certainly wasn’t transparent.  In my opinion, FR technology has been operating in the shadows.

The Toronto Police Service is only the second police force in Canada using facial recognition technology.  The Calgary Police Service began using facial recognition technology in 2014, and in largely the same way the TPS presently uses it: to match images of crime suspects taken from public and private cameras against its internal database of mugshots.  The single database the TPS relies upon comprises an astounding 1.5 million photographs. [FN1]

There is no doubt that the technology can be a valuable investigative tool: able to enhance the ability of police to identify criminals, to solve previously unsolved crimes, and to conclude lengthy and large-scale investigations with fewer resources. For those reasons, wider adoption of the technology by police seems inevitable. However, that does not mean that those who are using the technology should be left to determine whether or not the public is suitably protected in their use of it. Ensuring an appropriate framework that maximizes the benefits of FR and that prevents and minimizes its risks is no easy task. This  is the purview of legislators.  The core competencies of police lie elsewhere. 

And legislators are late to the party. The technology is already in wide use in the US but tends to be regulated through broad legal norms.  Various US government agencies (such as the FBI and the Department of Homeland Security) have implemented and developed their own protocols for its use.  Only three US states (Illinois, Texas, and Washington) have specific legislation dealing with the collection and use biometric information (which includes facial metrics).  Canada has no comparable legislation.

The lack of a robust legislative framework to govern the use of FR has resulted in the city of San Francisco, itself a US technology hub, recently imposing a prohibition on the use of facial recognition technology by its municipal departments and agencies. [FN2]

FR does not operate in a lawless world, but the lack of regulation of FR is concerning.  Consider that Canada lags behind the US in regulating the collection and use and dissemination of biometric information.  On the bright side, our federal House of Commons Standing Committee on Access to Information, Privacy and Ethics appears to be leading the way towards the development of informed policy-driven privacy legislation suitable to the times we live in.  But until new law is enacted to fill the gaps left by PIPEDA and the Privacy Act, for instance, we are left hoping.

There are many issues around the TPS's use of FR, that to my knowledge have not been adequately investigated. So as not to turn this particular blog into a treatise, I will deal briefly with only one: the fact that the key members of the Toronto Police FR team received some of their initial FR training from the Federal Bureau of Investigation (FBI). [FN3]  We should not be placated by the fact that Toronto police were FBI trained. 

The FBI has hardly been an exemplar in its use of FR technology. As US Congressman Jim  Jordan (a Ranking Member of the U.S. House Committee on Oversight and Reform) asked of the FBI on June 4, 2019,  “How are we supposed to have confidence, if you [the FBI] didn’t follow the rules when you set the thing [facial recognition system] up in the first place?"

It appears not to be the practice of the FBI to notify those who have been criminally charged that facial recognition has been used in their investigation. [FN4]. This is constitutionally problematic.  In Canada, the failure of police to disclose to an accused relevant information pertaining to his/her case violates the accused’s right to make full answer and defence under section 7 of the Canadian Charter of Rights and Freedoms.

Additionally, according to a letter from Comptroller General of the United States Gene L. Dodaro to Attorney General William P. Bar, “A series of six recommendations made by the U.S. Government Accountability Office (GAO) to the FBI in May, 2016 to ensure the privacy protections and accuracy of its facial recognition capabilities have yet to be followed.” [FN5]

For instance, the FBI has not taken steps to verify that the Next Generation Identification-Interstate Photo System (NGI-IPS), which contains approximately 30 million images, is sufficiently accurate. The FBI’s accuracy testing does not test for a false positive rate, and thus does not present a complete view of the system’s accuracy.  Further, the FBI privacy impact assessments and System of Records Notice (SORN) were not being published by the FBI as required. [FN6].  

None of this leads inevitably to the inference that the Toronto police have been guilty of the same laxity as the FBI.  But it deserves the question. A question that is now at least a year overdue. 



[FN1] No other databases are utilized. The R.C.M.P. does not have a national mugshot database.

[FN2] Imposed May 2019. 

[FN3] TPS Facial Recognition System team received Face Comparison and Identification training at the FBI's Criminal Justice Information Services Divison in Clarksburg, West Virginia.  This training is consistent with the guidelines and recommendations outlined by the Facial Identification Scientific Working Group. 

[FN4]  Testimony of Ms. Del Greco (FBI), House of Representatives, House Oversight and Government Reform Committee, June 4, 2019:  https://www.youtube.com/watch?v=GBkd5AUprYs  at 59:33/2:53:45

[FN5] Chris Burt, BiometricUpdate.com, “FBI not following GAO recommendations for face biometrics but expanding iris recognition work”.  Apr 22, 2019. https://www.biometricupdate.com/201904/fbi-not-following-gao-recommendations-for-face-biometrics-but-expanding-iris-recognition-work

[FN6] For the complete list of GAO recommendations see https://www.gao.gov/assets/700/698610.pdf

Comments

Popular posts from this blog

Warrantless Drug Searches (Section 11(7) of the CDSA)

Arrested at Home: Feeney Warrants

Night time Execution of a Search Warrant