A cyber project

[Apparently a slide in the presentation in the pitch mentioned below]

Corsight of reference 1 appear to be a leading company in the facial recognition business. Based in Israel. They can do things like telling a casino that the chap in a flashy blazer on the big roulette table is on their black list. Even, on a good day, if he is wearing a COVID face mask.

They have apparently been pitching to a conference put on by the Los Angeles investment bankers at reference 2 about how they think they can build systems to put a face to a DNA sample. So you get a DNA sample from a crime scene, they turn that into a face, you run that through you database of criminal faces and you get your criminal. The front cover of the brochure put out by the helpful bankers is snapped above. They do not, however, put out the pitch: the snap at the top of this post just leaked.

While Corsight just happen to have a couple of very senior US security types (presumably retired) on their advisory board.

They also seem to be mixed up with the Japanese company at reference 3, but I have not been able to work that one out. But whatever it is, it includes the handy presentation about Corsight at reference 4.

It seems that lots of people don’t believe that this can be done. Sex, colour of skin and colour of eyes, yes. Appearance of nose, no. At least, not yet.

It also seems that lots of people worry about privacy and human rights issues. Inter alia, they point to the difficulty that facial recognition systems have with very dark skins.

The security industry seems to be what pushes facial recognition along. While both the security industry and the ancestry industries are players in the world of genes. And it seems that, given the amount of DNA out there for medical research purposes linked to basic demographics like age and region of residence and the amount of DNA out there for ancestry purposes linked to family names, aka surnames, that it is now possible to break the anonymity of the first of these using data from the second. See, for example, reference 5. Not that this would particularly worry me. So some bad person got hold of the computer version of my DNA. So what? It is not as if they can use it to break into my bank account, at least not yet. But perhaps I ought to worry.

While I do worry about using DNA for criminal justice purposes. It is all very well talking about something based on tricky science that a non-specialist can’t hope to understand being more or less probable, but it is quite another matter saying that it meets the beyond reasonable doubt test of the public part of most criminal justice systems. At least you can see fingerprints and more or less understand what a fingerprint expert is saying; but with DNA and DNA experts this does not fly at all.

I have learned about things called short tandem repeats. It seems that the non-coding part of human DNA contains lots of these repeats, things like ‘TATATATATA’ and ‘GTCGTCGTCGTCGTCGTCGTCGTCGTCGTC’, with, in the usual way, A being Adenine, G Guanine, C Cytosine, and T Thymine. The number of these repeats is inherited, but also varies a good deal from person to unrelated person. Which makes them useful for identification purposes. What they are useful for otherwise is not clear at all, at least not to me.

And about the glossy but curiously uninformative website at reference 6. But which looks like a virtual health club for the rich, one which has succeeded in sucking in a lot of money.

This complicated yarn being kicked off by the article in the MIT technology review detailed at reference 7.

References

Reference 1: https://www.corsight.ai/. 

Reference 2: https://www.imperialcapital.com/. 

Reference 3: https://ist-security.com/.

Reference 4: https://ist-security.com/ist-wp/wp-content/uploads/2020/10/Corsight-Deck-2020Rev1.pdf.

Reference 5: Identifying Personal Genomes by Surname Inference – Melissa Gymrek, Amy L. McGuire, David Golan, Eran Halperin, Yaniv Erlich – 2013.

Reference 6: https://humanlongevity.com/. 

Reference 7: This company says it’s developing a system that can recognize your face from just your DNA: Though it almost certainly won’t work, it is a telling sign of where the field is heading – Tate Ryan-Mosley, MIT Technology Review – 2022.