Technologies offered to defeat facial recognition and provide alternative for tracking

surveillance-graffiti

The proliferation of facial recognition technology is giving rise to methods of subverting it, such as the laser pointers used by protestors in Hong Kong to avoid identification by facial recognition-enabled CCTV cameras, The Economist reports.

The machine learning techniques which power facial recognition algorithms are nothing like human vision, The Economist points out, and are therefore vulnerable to methods of disguise or deception which do not work on people.

New York University researcher Adam Harvey, who created the “cv Dazzle” make-up design method to fool facial recognition algorithms, has developed a line of clothing designed to trigger false positives and lower the confidence score for the wearer’s face, called HyperFace. Last year researchers developed a method for generating false negatives, and in theory even false positives, using infrared LEDs fastened to a baseball cap. Adversarial machine learning techniques have also yielded some images which are effective for fooling automatic facial recognition system.

Computer vision startup Traces AI has developed a method of tracking people with cameras while blurring out all faces in the frame, TechCrunch reports.

The company is participating in the latest round of Y Combinator’s accelerator program, and its founders believe facial recognition is too invasive of public privacy, according to the report. Traces AI’s technology tracks people based on their clothes, hairstyle, and other features. An advantage, according to the company, is that it allows searches to be conducted based on descriptions, rather than a photograph.

Whether the public would consider it any less invasive remains to be seen.

Related Posts

Article Topics

 |   |   |   |   |   |   |   |   | 

Comments

Leave a Reply

This site uses Akismet to reduce spam. Learn how your comment data is processed.

Most Read This Week

Featured Company

Biometrics Research

Biometrics White Papers

Biometrics Events

Explaining Biometrics