That old razzle-dazzle

[davywavy]

The purpose of camoflage has always been to break up shape and to disguise it against the background. However, back in the first world war, navies found they couldn't find an effective camoflage colouration to hide a ship against the variety of possible sea and sky shades. This meant that with advances in technology in naval rangefinding and targetting ships were often sitting ducks.

In response to this, the navy came up with Dazzle Camoflage. Rather than painting ships blue or grey to get them to blend in with the sea or sky, they were painted with bright, jumbled colours to break up their shape and confuse the eye of the beholder in order to make it harder to judge range or speed. So the war ended up with Battleships looking like this:



It's not known how effective Dazzle Camoflage was as a defense, but it was good for crew morale which the navy was well pleased with.

Anyway, as token_limey pointed out in the replies to my post yesterday, these same principles are now being applied to humans to defeat facial recognition software on CCTV, in with people it does work. You can see some examples here: http://cvdazzle.com/

Hoodies are so popular among teenagers because they hide the face (what's that you say? We spend all day every day filming people and then act surprised they hide their faces? Well, well), but I can really see the looks demonstrated on that website catching on. Would anybody be surprised to see people who looked like that getting on the tube? I wouldn't.

Comments

[crocodilewings.livejournal.com]

It's worth pointing out the distinction between facial recognition in the sense of telling whether an object is a face, and facial recognition in the sense of identifying whether two faces are the same. Confounding the former will probably confound the latter as well, but the former is an objectively easier task, so I'd expect the technology to improve quite quickly.

[davywavy.livejournal.com]

I'd still be prepared to make a small wager that a band made up of people who look like the ones on that site will bother the charts within the next 12 months.

Hang on, HANG ON. I think I've just explained Lady Gaga.

[hairyears.livejournal.com]

Interesting...

Google for more images (I can't, I'm on a mobile phone) of warships: one of the most effective tricks was the outline of a more distant ship - you can't hide a battleship by 'breaking up the outline' but you can throw off the rangefinding calculations.

In terms of faces and surveillance, this is a better bet than slashes of diagonal makeup, 'breaking up the outline' of the face but making the wearer far more conspicuous - the 'rangefinder' you're trying to fool is actually facial-recognition algorithms capturing the ratios of eyes and nose and mouth; I suspect than women with a subtle bias in their eyeliner and their lipstick can appear, to cameras, to be a completely different person.


Men might have an issue with that: I'm not so rash as to predict a future in which delinquent teenagers in Tottenham take make-up tips from London's Goth community.


Now consider natural mimics in nature: specifically, insects that have eye spots. They aren't photographically-perfect mimics; they exaggerate key features used by the primitive image-processing algorithms of their predators. So the next must-have accessory for Yoof will be facial adornments and hat badges imitating eyes: and by 'imitating' they need only be a pair of dots to fool the software.

In short: there are very simple patterns and disruptive 'mimic' objects that would pass unnoticed by human observers (and the users do not want to be conspicuous) that would be just effective as a Hallowe'en mask or a garish make-up job.