The FBI’s facial recognition program fails 20% of the time, according to a 2010 report recently released. But that’s not stopping the agency from moving forward with the project.
The report was obtained by EPIC (Electronic Privacy Information Center) through a Freedom of Information Act request. Regarding the FBI’s facial-recognition technology, the report says it “shall return an incorrect candidate a maximum of 20% of the time,” and, “when used against a searchable repository (i.e. faces looking directly into the camera; clear, no obstructions), it shall return the correct candidate a minimum of 85% of the time.”
EPIC has been tracking the FBI’s research programs over the last few years, utilizing the Freedom of Information act every chance they get in order to inform the public on what the agency is up to, even suing the agency earlier this year for failure to produce certain legally requested documents. And many claim that keeping the FBI and their programs under a microscope does the public justice, seeing as, “An innocent person may become part of an investigation because the technology isn’t completely accurate,” says Jeramie Scott, an attorney for EPIC who reviewed the documents.
Facial recognition technology has been dramatically improving over the last few years. But as the technology gets better, the public is left wondering how their privacy rights will be affected. This past June, we reported how the Statue of Liberty was now using the software to monitor its visitors. Meanwhile, Facebook has mainstreamed the technology with their recent “auto-tag” feature, and even former-funny-man-turned-Senator Al Franken recently expressed concern during a Senate Judiciary hearing, claiming facial recognition technology could be used to violate a person’s privacy without their knowledge.
Last year, FBI spokesman Jerome Pender told Congress that the bureau was planning to update its 2008 Privacy Impact Assessment, which would address “all evolutionary changes” in the originally-planned use of facial-recognition technology. However, even in the wake of the NSA surveillance scandal, the FBI has yet to release this “revised” assessment.
From a forensic video perspective, it would seem the FBI still has a ways to go before they can utilize facial recognition technology with standard video surveillance. After all, if their system fails 15%-20% of the time, even when matching pictures of faces taken in ideal settings, then safe to say the system is going to struggle matching faces seen in surveillance footage, especially as most surveillance cameras are mounted in elevated positions and often capture images and faces of people from off-angles. The failure rate will lower over time, but for now, automated facial recognition is not ready for integration with any government or private security system. Facial recognition will become a more reliable technology and useful tool to law enforcement and Homeland Security within a few years, but the question will remain: How do we protect our freedom and privacy as this occurs?