2024 Update on DHS’s Use of Face Recognition & Face Capture Technologies

The DHS Science and Technology Directorate (S&T) oversees testing and evaluation for FR/FC technologies. Testing happens both before technologies are put in the field where you might interact with them and at least every three years during operational use.

The Maryland Test Facility conducts robust, independent testing and evaluation of FR/FC technologies. Volunteers who represent a range of demographics sign up to help test biometric technologies in a lab set up to resemble real world conditions.

In addition to these performance tests, DHS program teams also work to continuously improve services to the public to ensure that you are having the best possible experience while DHS ensures national security.

DHS completed performance reviews of several priority uses of FR/FC, based on direct testing, analysis of operational reporting statistics, and reviews of third-party testing results. We analyzed demographic differentials where possible. Through this testing, we learned that:

• Overall, FR/FC systems performed extremely well for diverse demographic groups. On average, the technology worked more than 99% of the time for systems that are fully operational, like ID checks for travelers at the airport and ports of entry to the U.S.
• TSA Credential Authentication Technology, which is used to verify authenticity of IDs and flight reservations at TSA checkpoints, had no performance issues across any demographic group.
  • Our testing revealed an important finding for TSA PreCheck's prototype Touchless Identity Solution. While the face matching worked well, we encountered issues with the face detection algorithm used to verify if a photo contains a face before matching. This algorithm was accurate 88% to 97% of the time, with performance varying based on skin tone and self-reported race, gender, and age. To address this, TSA quickly introduced a manual photo capture step, which only adds 2-3 seconds to the process and does not affect the overall screening experience. TSA and DHS S&T are currently evaluating new algorithms to improve this step and plan to test and implement them later this year.

We noticed two other minor trends in test results that will be monitored going forward:

• For some CBP use cases, there were very small differences in measured face matching performance based on skin tone and self-reported race and age, ranging from less than 1% to 2-3%. Face matching still performed well overall, and the lowest success rate for any demographic group was 97%. This round of testing was only designed to reliably detect differences of 5% or greater, so we can’t say if smaller measured differences reflect true underlying differences in performance. We will continue to monitor these trends, refine our testing practices, and take action as appropriate.
• People interacted with systems quickly – in seconds, not minutes. The time to move through the FR/FC process ranged from less than 10 seconds for a Global Entry Touchless Portal to less than 30 seconds to do an ID check at an airport security checkpoint. In general, it took a few seconds longer for those aged 61 years and over to complete FR/FC interactions than those 60 years and under. While not a cause for immediate concern, we will work to ensure our systems continue to be usable across all age groups.

You can view detailed test results in the full report.