Sunday, October 21, 2018

Big data is face-recognizing you


At some point in the near future the harvesting of facial recognition images will replace the harvesting of Facebook profiles as the new technology cause célèbre. This prediction is prompted by the following chilling assertion from George Thaw, the CEO of scanning solutions developer Fuel3D:
3D facial recognition technology has traditionally been used in security, but thanks to advancements in technology and machine learning algorithms it now has massive potential to disrupt the retail industry and set brands apart by redefining customer experience.
Fuel3D's core technology platform was developed initially for medical imaging. Like the notorious Cambridge Analytica, Fuel3D's proprietary technology is a product of Oxbridge academia. Its BioVolume oncology application was developed at Oxford University more than a decade ago, and the company's registered office is at Oxford Science Park with another office at Greenville, North Carolina. The filed accounts state that the principal activity Fuel3D is the development, sale and application of 3D imaging technology. There are no grounds whatsoever to suggest that Fuel3D is motivated by anything than the standard ethical commercial ambitions and, doubtless, numerous other companies are developing similar technology. However this should not preclude using Fuel3D as a case study to raise generic concerns about the retail application of facial recognition technologies.

As pointed out in the quote, facial recognition had its earliest applications in the security industry. Two examples are EU passports, which use facial recognition to compare a scan of the subject's face with a photo held on the passport's chip, and the Apple iPhone X and Samsung Galaxy phones which have face ID security. Fuel3D's website states that it is developing applications for the retail sector encompassing personalised cosmetics and customised skincare, performance sports eyewear and head protection, custom footwear and clothing, and custom eyewear and virtual try-on. All these applications require a high resolution and extremely accurate three dimensional image of the client's face to be captured at point of sale. A Fuel3D video on YouTube convincingly demonstrates how these images can be captured very easily and quickly, while another video demonstrates how a facial image from a Fuel3D application can be printed to produce a very accurate 3D model of the subject's face.

The resolution of the images captured by facial recognition technology far exceeds the relatively crude photos used in biometric passports and other forms of photo ID. We now live in an age where the value of data is far greater than that of the software used to capture and store it. Facebook's commitment to facial recognition is widely known, and we have seen the value and power of Facebook's woolly personal profile data. Given the porous boundaries between commerce and security, the value of a data set linking Facebook's profile data to highly accurate three dimensional facial images captured in the retail environment is very obvious. And it gets worse: because another major player in the field is Amazon whose Rekognition application can, to quote their website "detect, analyze, and compare faces for a wide variety of user verification, people counting, and public safety use cases". Widespread concern has been expressed about the supply of Amazon Rekognition technology to enforcement agencies.

Encouraging the public to voluntarily provide a highly accurate face image at point of sale takes the danger of the misuse of facial recognition technology to a new and much higher level. Anyone thinking that this is just 1984-style scaremongering is referred to a recent Guardian article which reports a psychologist as claiming that using facial recognition software, artificial intelligence can detect a person's sexuality and politics. The article explains how psychologist Michal Kosinski published a controversial paper showing that face-analysing algorithms could distinguish between photographs of gay and straight people, and goes on to tell how his work aroused interest within the Russian government. Paths converge here as Michal Kosinski's research was undertaken at Cambridge University and has links to the infamous Cambridge Analytica data harvesting. Given recent allegations of electoral interference by Russia, the possibility of a data set of merged Facebook profile and high resolution facial images being analysed by AI to interpolate political allegiance is very disturbing.

Our obsessive love affair with new technology means Facebook's user base has remained remarkably resilient despite several recent high profile exposés of data abuse. So it is certain that facial recognition applications will be avidly embraced as part of our retail therapy, and once again a new technology will be presumed innocent regardless of any evidence to the contrary. But beware; because big data is face-recognizing you.

Header image is from an Andy Tran YouTube video. My social media accounts are deleted. But new Overgrown Path posts are available via RSS/email by entering your email address in the right-hand sidebar. Any copyrighted material is included for critical analysis, and will be removed at the request of copyright owner(s).

No comments: