Facial recognition has many applications. One of them is law enforcement. This article looks at the merits of facial recognition in law enforcement and how responsible use can make it even more effective in fighting crime.
has many applications. One of them is law enforcement
. This article looks at the merits of facial recognition in law enforcement and how responsible use can make the technology even more effective in fighting crime.
Facial recognition is increasingly used in law enforcement. In the U.S., for example, the technology is employed by agencies on the federal, state and local levels (the only exceptions being Vermont and 19 cities and towns that have bans in place prohibiting the use of facial recognition in law enforcement). Increased adoption is due to its various benefits.
“Without facial recognition, investigators are left with manual review of photos in criminal record databases and open-source information, as well as requests for public help with identifying persons in the photo. Facial recognition can aid and hasten this process by comparing such photos with those in available databases and returning photos with high similarity scores,” said Jake Parker, Senior Director of Government Relations at Security Industry Association (SIA).
How it works
It is important to note that in the U.S., facial recognition remains limited to a post-incident investigative tool to generate leads and aid identification, but not confirm identities. It works by way of matching the image of a person in question against images in a database.
“The image could be from any available source that provides adequate quality for comparison, such as security camera footage and cell phone cameras. This photo is compared against an available database of images using facial recognition software, which returns all or a set ranking of potential matches over a preset threshold similarity score. Personnel then determine whether any returned matches represent leads that should be investigated further,” Parker said.
He added: “The most common data sources queried are state or regional repositories of arrest records and associated photos, sometimes with special units at those levels that use the technology and field requests. Some agencies will also use other facial recognition tools that provide the ability query open-source information on the internet, from mugshots to news articles and social media posts. All this data is considered public information in the U.S., though there has been some controversy regarding the use of online photos.”
Effectiveness and success stories
Indeed, there are merits associated with facial recognition in law enforcement. These are reflected in various success stories cited by Parker.
“The New York City Police Department reported having conducted 9,850 searches of its criminal records in 2019 alone, resulting in 2,510 possible matches related to 68 murders, 66 rapes, 277 felony assaults, 386 robberies and 525 grand larcenies in that year and no known related misidentification or mistaken arrests,” he said. “In 2019, the FBI reported it had conducted 390,186 searches on a variety of government databases in support of active FBI investigations and 152,565 on behalf of other law enforcement agencies between 2017 and 2019. In just one example, the FBI used the technology in 2017 to help identify and ultimately apprehend a child abductor and sexual predator in Oregon after a 16-year manhunt.”
Inevitably, there are concerns towards face recognition in law enforcement due to privacy
and accuracy issues. Concerns that the technology has bias against certain ethnic groups, for example, have been voiced by critics. Yet according to Parker, these concerns have become less and less of an issue due to progress in facial recognition over the years.
“In a more recent look at [National Institute of Standards and Technology (NIST)]’s ongoing evaluations of facial recognition technology from July 2022 – under testing that most closely simulates investigative applications – the top 30 algorithms were able to retrieve matching photos at the highest ranked score from a database of over 12 million mugshot images, with ‘miss rates’ of less than 2 percent,” Parker said. “The July NIST verification test data generally shows the top 150 algorithms are at least 99 percent accurate across black male, white male, black female and white female demographics using such images.”
He added: “Without minimizing the negative effects of mistaken arrest for the affected individuals and their families, concerns about law enforcement use of facial recognition must be balanced against its proven benefits and we should consider whether more misidentifications would result without it.”
Shifting public opinion
In fact, Parker cites the trend in the U.S. where more people are supportive
of law enforcement usage of facial recognition.
“All of this is consistent with the most comprehensive public opinion research to date on Americans’ views of facial recognition technology, commissioned by SIA, which found that 68 percent of Americans believe facial recognition can make society safer, 66 percent believe law enforcement’s use is appropriate and 57 percent were even comfortable with including their images in a database searchable using the technology if it’s for a public safety purpose,” Parker said.
According to Parker, there is growing consensus among law enforcement professionals on the necessity of facial recognition as well as appropriate processes and rules surrounding its use.
“This is indicated by this year’s publication of guidance from both the Major Cities Chiefs Association and the International Association of Chiefs of Police. Many of these guidelines are reflected in common elements of new laws and policies we have seen so far, such as addressing concerns about possible misidentification by ensuring potential match results alone are never considered probable cause for an arrest or to obtain a search warrant,” he said. “This also includes requirements like multiple levels of supervisor review prior to performing facial recognition comparisons, as well as documentation and tracking of details such as the source of the comparison image and underlying case information.”
He adds while SIA has documented many success stories, it’s the association’s belief that facial recognition must be accompanied by the right use policies and requirements. “Setting parameters for law enforcement applications can be done in a way that ensures it is used in bounded, appropriate ways that benefit society and are constitutionally sound,” he said.