Corsight’s Upcoming DNA to FACE: 'Terrifying' Warns Privacy Expert

By Donald Maye, Published Jan 31, 2022, 07:52am EST (Info+)

Corsight plans to release a new product that combines DNA and face recognition technology and could have significant law enforcement and privacy implications.

IPVM Image

In this report, we examine Corsight’s product roadmap for "DNA to FACE," presented at the 2021 Imperial Capital Investors Conference, possible use cases for the technology, and warnings from a privacy expert.

IPVM collaborated with MIT Technology Review on this report, see the MIT Technology Review article: This company says it’s developing a system that can recognize your face from just your DNA

Corsight’s Roadmap - DNA to FACE Presented

On December 15th, IPVM attended a presentation by Corsight's CEO, Robert Watts, and EVP of Business Development, Ofer Ronen, at the Imperial Capital Investors Conference in New York.

Corsight's new product roadmap, shown below, included three products: "VOICE to FACE," "DNA to FACE," and "MOVEMENT".

IPVM Image

In the slide, Corsight describes "DNA to FACE" as:

Constructs a physical profile by analyzing genetic material collected in a DNA sample

Prior to this presentation, IPVM was unaware of a company attempting to commercialize a face recognition product associated with a DNA sample.

Corsight did not provide details on how "VOICE to FACE" would work. The company's description of "MOVEMENT" is more commonly known as gait recognition. IPVM previously examined this analytic and its limitations.

"Company Confidential" - Confirms Intent to Launch

IPVM asked Corsight about "DNA to FACE" and its intended use. Watts replied to IPVM and, citing confidentiality concerns, declined to discuss "DNA to FACE" details. However, the CEO confirmed that Corsight intends to launch the product.

We are not engaging with the press at the moment as the details of what we are doing are company confidential. We will release information once the launch is announced.

Privacy Expert: Technology "Terrifying"

IPVM Image

IPVM and MIT Technology Review spoke with privacy expert Albert Fox Cahn, founder of the Surveillance Technology Oversight Project. Cahn raised concerns of potential bias and capacity for error associated with combining DNA with face recognition technology, calling the combination 'terrifying':

The idea that you're now gonna try to combine the two technologies is really terrifying because not just because of the obvious potential drive of bias involved, but also the capacity for error here.

Cahn also questioned the feasibility of such technology:

The idea you're gonna be able to run a facial recognition search against a composite image and have anything approaching reliable results to me is just a staggering claim, it just seems inconceivable.

And so the idea that you're gonna be able to create something with the level of granularity and fidelity, that's necessary to run a face match search to me, that's preposterous, that is pseudoscience.

Possible Use Cases for DNA to FACE

While Corsight declined to elaborate on specific use cases, EVP Ofer Ronen stated during Corsight’s presentation that the progression of AI and machine learning have made this technology inevitable.

One of the plausible use case for combining the two technologies is law enforcement and forensic analysis of crime scenes. For example, creating a composite image of a suspect from DNA found at a crime scene would enable investigators to match the image to existing photo databases or show to witnesses. Conceptually, such technology could also be used to analyze cold cases involving DNA evidence.

Outlook

Corsight risks significant resistance to the commercialization of "DNA to FACE." Because the specific use cases are not clear and the accuracy of Corsight's technology is unknown. If the product is eventually released, as Corsight intends, "DNA to FACE" will draw more scrutiny from law enforcement/legal and privacy-rights perspectives.

1 report cite this report:

ISC West 2022 Show Report on Mar 25, 2022
Not since 2019 has ISC West felt "normal". 2020 was canceled, 2021 was held...

Comments (7)

Only IPVM Subscribers may comment. Login or Join.

While I do not know what Corsight's intentions are with DNA claims, this fits with the company's track record of making aggressive claims. For example, see the red circle drawn around a claimed facial recognition match in the company's marketing below:

IPVM Image

From their comments to us, they genuinely believe they can reliably and accurately do these things, like above where they recognize a face ("Ray Williams") in a "low resolution camera", with a 30+ meter Field of View, of a person wearing a mask at a bad angle to the camera.

I think investors will find the claim interesting, so there is logic to pitch this at an investor conference but I'd be skeptical about this becoming a successful commercial offering for them.

Agree: 3
Disagree
Informative: 3
Unhelpful
Funny

I definitely feel that companies like this, once the general public learns about how invasive this is, will start to wither and fade. This is why I feel that Americans should demand a digital bill of rights that includes a right to privacy and to be forgotten from the internet.

Agree: 4
Disagree: 1
Informative
Unhelpful
Funny

Some of the history of this tech is highlighted in MIT Technology Review's piece:

Corsight’s idea is not entirely new. Human Longevity, a “genomics-based, health intelligence” company founded by Silicon Valley celebrities Craig Venter and Peter Diamandis, claimed to have used DNA to predict faces in 2017. MIT Technology Review reported then that experts, however, were doubtful. A former employee of Human Longevity said the company can’t pick a person out of a crowd using a genome, and Yaniv Erlich, chief science officer of the genealogy platform MyHeritage, published a response laying out major flaws in the research.

2019 research published in Nature Communications: Facial recognition from DNA using face-to-DNA classifiers describes such a methodology:

IPVM Image

2017 research published in PNAS: Identification of individuals by trait prediction using whole-genome sequencing data noted that traits such as gender, skin color and eye color were predicted with high accuracy. However, more complex traits were less accurate.

We predicted genetically simple traits such as eye color, skin color, and sex at high accuracy. However, for complex traits, our models explained only small fractions of the observed phenotypic variation.

Agree
Disagree
Informative: 5
Unhelpful
Funny

Corsight describes "DNA to FACE" as: "Constructs a physical profile by analyzing genetic material collected in a DNA sample" [empasis added]

This strikes me as a very bad form of profiling because how much information specific to an individual's face are they going to get from DNA? From the above research, it appears very little.

We predicted genetically simple traits such as eye color, skin color, and sex at high accuracy. However, for complex traits, our models explained only small fractions of the observed phenotypic variation.

That's not anything close to actual facial recognition as it is normally used, i.e., recognizing a specific face, not guessing at skin color and sex.

Agree
Disagree
Informative
Unhelpful
Funny

A very interesting comment there.

While I might find it plausible to be able to identify certain, specific, markers for possible identification from DNA, eye, hair and skin colour for example spring to mind and, to a degree, phenotypical presentation. But the idea that you could use DNA to reconstruct what a face would look like? When so many aspects of how DNA presents itself in a child and adult? This sounds like the worst sort of pseudo-science fiction.

Agree
Disagree
Informative: 1
Unhelpful
Funny

Glad that IPVM brings attention to this. It certainly belongs to an implausible research category (just look at 23andme predictions out of DNA - they merely predict increased couple of percent likelihood of only a dozen most common genetic diseases, even though they have MLNs of records for the AI through). But the scary thing is that they still may get investor and a desperate customer interest, as has been the case with several other Israeli companies, such as the one predicting criminal traits and terrorism intent from a photo!

Agree
Disagree
Informative
Unhelpful
Funny

Taking a guess at how this could be done, I would assume you could reverse the look up process. Today DNA is used with undeniable accuracy for relative and paternity confirmation. So theoretically by getting DNA from a crime scene you could perhaps do a reverse look up in a DNA database for relatives of the suspect (assuming you have access to such databases) and provide a pretty good guess of what they might look like based on nearest or immediate relatives, especially if several key markers (skin/eye colour, sex,etc) are already provided through the initial DNA testing. It doesn't necessarily guarantee accuracy all the time, but it is a start to somewhere in the investigation. The consequence of such capabilities would indeed be terrible and one of the last nails in the privacy coffin...

Agree
Disagree
Informative
Unhelpful
Funny
Loading Related Reports