Heat From Fingertips
AgitatedRiot
Posts: 4,589
Heat From Fingertips, Ai is a big game changer. A device created where the warmth from your fingertips, with the help of AI, has been developed to crack passwords on computer systems. I read the article. I type a lot. I don't know how they can do it.
Heat from fingertips can be used to crack passwords, researchers find (yahoo.com)
Daz 3D is part of 
Connect
DAZ Productions, Inc.
7533 S Center View Ct #4664
West Jordan, UT 84084
Licensing Agreement | Terms of Service | Privacy Policy | EULA
© 2025 Daz Productions Inc. All Rights Reserved.
Comments
ATMs too I guess
I guess.
This is by no means new. Infrared cameras have been used for the number pads of cach machines quite a while ago.
AI likely helps interpreting the results, and makes the method more reliable, possibly cheaper to use. Infrared, radar,... all not new, all work on keyboards, in different scenarios.
Accidental comment: So maybe you've heard of post-production haters? They usually tap wildly onto all keys on public keypads, just to screw up all post production.
Since the beginning of the article "Heat-detecting cameras," it doesn't say they are Infrared. However, some may interpret that as infrared.
Heat, infrared, "thermal", different wavelengths, reflective properties... same point. It's been used by services for a longer time. Machine learning may push the edge of the doable by a good bit, but it basically has never been safe not to wipe all traces, including heat, in public place ;).
Elaborate: The machine learning system mentioned in the article seems to help finding the order of keys pressed better. This may be a significant improvement, but maybe it's not new per se either, because it's based on physics which sensors can detect, e.g. because earlier pressed keys should be cooler. It could also be, that detection becomes very fast and cheap with machine learning, reducing the necessary expertise to a handheld device, enabling more ordinary criminals to access the technique. Maybe such can work with some smartphones for very recent keyboard use. If you have been weary about gone-wrong prosecuters or governmental agencies, though, this is not a new issue :).
Infrared cameras have already been used for skimming with cache-machines, so it's a question of how far they can push it, to analyse images from keyboards, taken by heat-sensitive sensor technology. An interesting point for defence would be, if you not only have to use 32 differing random generated digits, just to be on the safe side, but have to add redundancy, to obduscate the key pressing order, blowing up password sizes to nowhere. Or you always keep playing the cat on the keyboard... but seriously: typing passwords in public probably is not wise (wasn't really before, but with this technique it's going to be near-suicidal). Someone walks up, you have typed your password, now you have to cover the keyboard, so they don't take an image of it, ridiculous...
Elaborate for pin-based devices: Some smartphones offer to shuffle the key positions for unlocking the phone, and there are also researched concepts for shuffling inputs on cache-machines, where you do multiple runs with fault-tolerance built-in.
Oh hay I've watched one program wherein one doesn't need anything fancy at all ... just some face powder from the lady's compact and that full brush to gently fill the prints with powder :-)
Maybe in the more classic period of spying, for real time operation, you'd have needed real-time access to experts to make the most of images taken and information gathered. Perhaps more has been possible, but other methods have just been cheaper, as well as available at all, due to the nature of "being intelligence services". Thinking of very small camera technology. Heat-sensitive sensors and radar have been available for quite a while, more broadly, though.
The machine-learning-based method has the advantage of delivering actual results, including probable order of keypresses (!), within a very short time frame, and for an outlook... possibly just from one image taken, even from some angle, enabling detached real-time operation. Never assume magic, though.
Machine learning may stumble here and there with imperfect training data, e.g. judging the success to be very low, because all images with low success had been on brightly colored keyboards, and stuff like that. It's just not magic. Remember red vs. green hydrants...