CIA Drones May Have Used Illegal, Inaccurate Code 279
skids writes "Coders hate having to rush code out the door before it's ready. They also hate it when the customer starts making unreasonable demands. What they hate even more is when the customer reverse engineers the product and starts selling their own inferior product. But what really ticks them off is when that buggy, knockoff product might be used by targeting systems in military unmanned drone attacks, and the bugs introduce location errors of up to 13 meters. That's what purportedly happened to software developer IISi, based on an ongoing boardroom/courtroom drama that will leave any hard-pressed coder appreciating just how much worse his job could get. The saddest part? The CIA assumed the bug was a feature. The tinfoil-hat-inducing part? The alleged perpetrators just got bought by IBM."
13 meters? (Score:5, Interesting)
so what?
hellfires are laser guided, not GPS. a predator reporting its position as being 13 meters wrong is basically nothing....and a non-issue with regards to missile targeting.
if the predator was dropping JDAMS, i could see the issue. but even then, 13 meters is well within the CPE allowed for the JDAM.
Lest we forget (Score:1, Interesting)
However, as I suspect that you're writing that from your parents' basement, I doubt that you actually know any history, or were even around for the Cold War."We are too easily impressed by small wars nowadays"- if you knew any history, you would know that the Western invasion of Germany was a limited war because high casualties would not be accepted by the American and British public. Read up on Eisenhower. You need to learn about the greatest American general.
Re:Let me see if I have this straight (Score:5, Interesting)
Sadly, I can top that story. I used to work for a government contractor that took blueprints and had them redrawn in AutoCAD in St. Petersburg, Russia.
Our main client was Los Alamos National Labs. We sent them the blueprints for almost every building there.
Re:In Soviet America, Zeroes Divide You! (Score:5, Interesting)
I read an interesting article in a mainstream magazine about Air Force drone pilots. Basically, they sit in Nevada and control drones in Afghanistan. I was expecting to read about how jaded and eager they were to press buttons at the drop of the hat, but what I found was the opposite.
The drones are capable of staying in the air for days at a time, monitoring a target (person). They have cameras running, and multiple shifts of human crews watching the video feeds and analyzing what's going on. In the process, they are able to ascertain with frightening accuracy that yes, this particular man is a terrorist: Here's the video feed of him buying some weapons, and here's the part X hours later where we just watched him create a roadside IED. Being able to keep someone under direct video surveillance (including thermal, if I recall correctly, so being indoors didn't help a lot) meant that for at least some targets, they were very sure that that person was a bad guy.
We have people halfway around the globe pushing buttons to kill people about whom we have reams of (video) evidence showing hostile behavior. I think that's better informed killing than having combat teams need to go in and do the same killing on foot, with potentially faulty intelligence, and without (at times) being able to mount multi-day uninterrupted surveillance.
Re:Wow. (Score:3, Interesting)
I'm guessing this is adaptable to alternative sensor platforms, not just cell towers. If they want that kind of accuracy from the SIGINT/DF hardware, they can probably get it. The problem is that they might not have a handle on the systematic errors being introduced into their targeting.
For example, say they slave the on station Predator optics to data from this software so they can pick up a guy in a town and follow him to wherever he's going. Everything is peachy, because they know there was nobody within 10m (but there were people 12m away) and the system's supposed to be accurate to 5m (or whatever). They have a good fix. CIA decides to make him an ex-person and maybe kills the wrong guy.
I hope this doesn't happen. I hope there are redundant checks within such a program to keep these things from happening. Maybe he has to make two phone calls. Who knows? The original contractor didn't know specifically how their software would be used. They wanted to ensure that the new hardware would match the old based on their regression testing so that as much as humanly possible, there would be no surprises.
Re:I think they buried the lead.... (Score:3, Interesting)
Re:Lest we forget (Score:2, Interesting)
Actually, Marshall jumped Eisenhower to the top because Eisenhower impressed the hell out of Marshall during 1941 war games in Louisiana when a lot of the show was a mess from career Regular Army clowns screwing up. Eisenhower got a lot of stuff done right despite the chaos and systemic cluster-fucks.
Marshall ruthlessly cleaned house after the mess and Eisenhower got jumped over a lot of RA types. People were pissed, but it kind of worked out OK, didn't it.
Re:Wow. (Score:3, Interesting)
True, but the error is additive, and any other errors in the pipeline stack up.
Pretend for a minute that the munition has a 6 meter kill zone. Say you have a tracking tech that's accurate to 10 meters. Likely anyone in that +/- 10m area is going to be very sorry even if they don't get dead right away.
Now, introduce another 10 meter inaccuracy. This means that you can be anywhere from bullseye to 20 meters away. The odds have suddenly gone from "most likely dead" to "more likely unharmed" - not a desired result! This is a recipe for failed strikes and collateral damage.