DARPA worried battlefield mixed reality vulnerable to 'cognitive attacks' DARPA is launching a program to head off "cognitive attacks" for mixed reality headsets that could, in theory, cripple future warfighters when deployed. The forthcoming Intrinsic Cognitive Security (ICS) program was sent out to tender this week, along with a proposers' day on October 20. On that day, DARPA officials will let …
Reminds of a passage from Star Wars Volume 3: The Last Command (Timothy Zahn). I type the excerpt below :
"I'm Councilor Organa Solo," Leia identified herself. "This is Ghent, an expert slicer. Can you use him ?"
"I don't know," the colonel said, throwing the kid a speculative look. "Ever tackled an Imperial battle encrypt code, Ghent ?"
"Nope," Ghent said. "Never seen one. I've sliced a couple of their regular military encrypts, though."
"Which ones ?"
Ghent's eyes went a little foggy. "Well, there was one called a Lepido program. Oh, and there was something called the ILKO encrypt back when I was twelve. That was a tough one - took me almost two months to slice."
Someone whistled softly. "Is that good ?" Leia asked.
The colonel snorted. "I'd say so, yes. ILKO was one of the master encrypt codes the Empire used for the data transfer between Coruscant and the original Death Star construction facility at Horuz. It took us nearly a month to crack it." He beckoned. "Come over here, son - we've got a console for you right here. If you liked ILKO, you're going to love battle encrypts."
Methinks DARPA and Uncle Sam's abiding military problem and increasingly obvious difficulty is in providing stupid enough combatants/grunts/cannon fodder whenever the worlds around them are awash with sensitive and much more believable alternative intelligence feeds which conflict with proven misleading and crooked self-serving establishment mainstream media tales/dodgy government news.
It is why that old and decrepit and rapidly failing public address and printed press system is so up in virtual arms against all of the new virtual social and anti-social media platforms so beloved of the young and dispossessed and the new generations of citizen broadbandcasters, branding them with all manner of misinformation and disinformation labels whenever all that is happening is the "official" narrative is not being believed because it is being increasingly easily proven to be manifestly false and deliberately fake.
This just feels like paving the way for claims about hacking when the troops start reporting that their expensive, stupid kit makes them ill for obvious reasons. Reminds me of the 'Havana syndrome' which was shown to be totally psychosomatic at best and just hangovers at worst, but years of made up stories about Chinese/Russian/insert-boogeyman scifi weapons and wasted money on investigations later...
This whole article is about attacks against Mixed Reality systems and has bugger all to do with anything you could sanely call "cognitive attacks"!
> Cognitive attacks go beyond just disrupting MR headsets and can also include actions like planting real-world objects to overwhelm displays, using physical objects to cause false alarms
No, that is not "going beyond" disrupting the MR system: like any other system, if you can feed it inputs that make it glitchy, it is being disrupted. Are we supposed to read "disrupt" as simply meaning cutting the wires, and everything else is "beyond" that?
> The core technical hypothesis of the program is that formal methods can be extended with cognitive guarantees and models to protect mixed reality users from cognitive attacks
You mean they are deploying software that HASN'T been subjected to formal methods? They have so long ago already given up on the idea of having robust avionics and fire control systems that they have to (re)introduce the concepts of formal methods just to get them applied to MR systems? Although, come to think of it, the current pushers of MR, AR and VR do strike one as never having heard of formal methods - barely heard of QA, some of them.
> In other words, DARPA wants to mathematically represent cognitive models of "human perception, action, memory and reasoning" and figure out how to build some form of universal protection against cognitive attacks
Ah, now that does sound like they are looking at something that deserves to be called "cognitive attacks"; maybe it will get juicy:
> "The ICS program ... as part of MR system development
Oh, nope, they are just talking about MR still. I was just getting excited at the prospect of dealing with some of the weirder ideas about using techniques from optical illusion and similar fields to befuddle the unaugmented human (think something between "dazzle ships" and trying to defictionalise the "fractal parrot"[1]).
So the "model of human cognition etc" can just be "keep to this framerate, don't rely on humans to remember anything ("whilst under stress?" "No, I mean - ok, let's go with that") and don't make the screen flash dark/bright at, ooh, 4Hz even if that *does* reflect the rate you are detecting bangs - and definitely not red!
> Given it is still early days for DARPA's ICS program, soldiers testing IVAS and other mixed reality devices won't be able to rely on cognitive security features anytime soon.
Given they have yet to solve the problem of lag and nausea in the first place, this is not a surprising result.
But, honestly, what is it they are actually asking for, under all the management speak? To start with, the use of formal methods to - guarantee that the frame rate is maintained to prevent lag? In other words, basic requirements of a Real Time system - "you have 10 milli seconds per frame: if you can't draw all the details in that time, skip them and update the display" instead of just doing *all* the work *every* time and just hope the nasty enemy hadn't put so many little walls onto the ground that you drop to 5 fps update[2]. BTW remember this is Mixed Reality, so you still get a view of the real world, which hopefully still manages a decent lag-free update.
> "Methods for developing protected MR systems will be developed by the ICS program before MR systems lacking protections are pervasive and essential,"
They are going to get some actual computer scientists, Old Geezers and others who actually *trained* in the Dork Arts to work on the MR headsets, not just gung-ho "visionaries" and tech-bros?
Well, if they *really* have to fling around such over the top language in order to get the grant money and the ability to hire the people they need, not just the ones they get, then, sigh, go with it. I guess at bottom it is no worse than any other grant proposal ("if you can just say the Rust compiler will investigate the effectiveness of Neural Nets, we can apply to this body as well - just keep Sheldon away from them, or he'll say we already know what the result of that investigation will be").
[1] David Langford, "BLIT" and "Different kinds of darkness" in particular.
[2] I am probably *very* out of date on game programming - which seems to be where the current crop of headset people are coming from (again) - but it seems that this "just do it all and fingers crossed" approach is what actually happens nowadays - which is why you get all these weird reviews about frame rates and "we got 182 fps from this GPU, but 184 from this one, so buy the latter". When the talk used[3] to be about games as Real Time and they either apologised and never ran or ran at 30fps, rock solid, didn't matter what extra cycles you threw at it, you got 30fps; the excess may be used to improve some shading or whatever.
[3] many moons ago and in discussions (we didn't have lectures about game writing back then) whilst marvelling at the latest hand-soldered MPU boards)
The Register 
Biting the hand that feeds IT
Copyright. All rights reserved © 1998–2025
