DHS Testing Shows Improved Biometric Performance on Masked Faces

The Department of Homeland Security (DHS) is working to improve its facial recognition capabilities when subjects are wearing face masks. The agency had previously expressed frustration with the widespread adoption of protective masks during the COVID-19 pandemic, with leaked documents revealing that the agency was worried that the masks would limit its ability to identify members of the public with facial recognition.

However, a recent controlled scenario test suggests that the agency is poised to overcome those limitations. The test was conducted in collaboration with the Science and Technology Directorate (S&T) over the course of 10 days at the latter’s 2020 Biometric Technology Rally, which took place at the agency’s Maryland Test Facility in the fall.

During the test, DHS evaluated 60 facial recognition configurations using six face and/or iris matching systems, as well as 10 separate matching algorithms. The technologies were applied to a group of 582 volunteers from 60 different countries, and were evaluated based on their masked and unmasked processing capabilities.

The results showed that while the facial recognition is still more accurate when subjects are not wearing masks, the best systems are able to identify masked individuals with a high level of consistency. Most notably, one system was able to recognize people with masks 96 percent of the time (the figure was closer to 100 percent for unmasked individuals).

Having said that, there was a considerable degree of variance between different facial recognition systems. The median identification rate was only 77 percent in the masked test, compared to a 93 percent median in the unmasked scenario.

Given the inconsistent results, the DHS would presumably need to conduct more tests to feel confident in its mask recognition setup. The agency is nevertheless hopeful that it will soon have a system that lets travelers pass through airport security checkpoints without removing their masks, which will in turn reduce the threat of diseases like COVID-19.

The Department of Homeland Security has come under fire for a controversial policy proposal that would give the agency the authority to collect much more biometric data from immigrants and US citizens. Meanwhile, organizations like the WEF have argued that technologies like facial recognition will help improve safety and restore confidence in air travel after the pandemic.

–

January 7, 2021 – by Eric Weiss