New facial recognition tools and defences for data and privacy protection
Several new tools have been developed to help consumers protect their privacy or data, either from or with facial recognition technology.
An app called DoNotPay, which charges $3 a month to take care of automated tasks like contesting parking tickets and cancelling free trials, has added a feature it calls ‘Photo Ninja,’ Input reports, which it says can render images unmatchable by facial recognition.
It does this through a combination of AI enhancement techniques like steganography, detection perturbation, and visible overlays.
TechCrunch Security Editor Zack Whittaker says on Twitter that the tool did not work when he tried it with Amazon Rekognition.
Waldo Photos, meanwhile, has been issued a patent for a tool that uses face biometrics to protect copyright on mobile proofs provided by professional photographers, according to a company announcement.
The platform’s FaceBlocker feature recognizes the face of potential purchasers and blocks their faces with a Waldo logo to prevent the image from being stolen with a screenshot.
Waldo says its platform is used by professional photographers for images from gymnastics meets, beaty pageants and schools, and the company says many of its early adopters have doubled their sales.
Private messaging service smashes crowdfunding goal
Yeo Messaging service, which uses continuous authentication through face biometrics to shield messages on the platform from unintended viewers, has raised £768,688 (just over US$1 million), or 157 percent of its target in a crowdfunding effort.
The messaging company says its continuous facial authentication technology is patent-protected. Other privacy-protection features offered by Yeo include geofencing and a ‘Burn After Reading’ tool. The service will be marketed to regulated industries like healthcare and fintech, with 4 versions of its app, including Pro, Business, and Enterprise, running on a software-as-a-service (SaaS) business model.
The company is offering investors equity at a pre-money valuation of £10.1 million ($14 million).
ICLR highlights facial recognition-blockers
A review of the new biometric application area by MIT Technology Review notes that this week’s International Conference on Learning Representations (ICLR) shows their increasing prominence.
LowKey, an adversarial AI-based tool, will be presented at ICLR.
Another tool, developed by researchers from Australia and China, seeks to introduce errors into images that prevent deep learning models from processing them with error-minimizing noise.
Predecessor software Fawkes has now been downloaded almost 500,000 times, and has been adapted into a web service by a third party.
In self-testing, Fawkes defeated 100 percent of attempted matches by cloud-based face biometrics algorithms, while LowKey’s developers found it reduced Rekognition and Microsoft Azure matches to below 1 percent.