As Amazon Face Recognition Tool Mismatches Politicians With Mugshots, Research Finds It Working Worse Than 2 Years Ago

News and Analysis

Tests performed by Comparitech showed that Amazon’s face recognition technology incorrectly matched over a hundred of the United States and United Kingdom politicians with photos of arrested people.

To perform the experiment, tech research company Comparitech made a selection of 1,959 publicly available images of lawmakers, wherein 530 politicians were from the U.S. and 1,429 politicians were from the U.K, according to a report released on May 28.

No Improvement?

The photos were matched against four sets of 25,000 randomly chosen images of arrested people via the Amazon Rekognition tool. In some cases, a single politician was incorrectly identified more than once against multiple mugshots.

Comparitech further pointed out the racial issue as Rekognition turned out to be racially biased. The report detailed:

“Out of the 12 politicians who were misidentified at a confidence threshold of 90 percent or higher, six were not white. That means half of the misidentified people were people of color, even though non-whites only make up about one-fifth of the U.S. Congress and one-tenth of the U.K. parliament.”

Out of 530 U.S. politicians, an average of 32 persons were mismatched to images in the arrest database, at the default 80% confidence threshold. When it came to the U.K. lawmakers, the system incorrectly matched an average of 73 persons out of 1,429, at the 80% confidence threshold.

Comparitech compared the results with those provided by the American Civil Liberties Union in 2018, and concluded that “by those standards, Amazon’s face recognition hasn’t improved and even performed worse than what the ACLU posited two years ago.”

Facial Recognition Market Size is Projected to Grow

According to a report from Component, the global facial recognition market size is expected to reach $7 billion by 2024. 3D facial recognition will reportedly have the highest market share over the forecast period.

“With the developments in 3D facial recognition software and services, the implementation of facial recognition is increasing, especially in healthcare IT solutions, payments, and commerce sectors. Hence, the 3D facial recognition segment is expected to gain a higher adoption rate across the globe,” the report said.

The more popular surveillance tools become the more concerns they raise. Recently, The Register reported about a security mishap on the part of an automatic number-plate recognition system run by the City Council of Sheffield, U.K.

It turned out that the system’s internal management dashboard wasn’t protected even by a password and could be accessed by entering its IP address in a browser. As a result, 8.6 million records were exposed, potentially allowing anyone to precisely deduct journeys of thousands of people down to a minute.

Written by Ana Alexandre

Follow us on Twitter and Facebook and join our Telegram channel to know what’s up with crypto and why it’s important.

Found a typo? Highlight text and press CTRL+ENTER

Subscribe to our Newsletter


Related posts

Tags: , ,