Don't judge BadBios without sufficient evidence
greebo.net
greebo.net
Article then claims professional researchers are saying:
> "Because I personally can’t verify this issue is true, the issue must be false. QED."
whereas I read it more as:
> "This is a really crazy, far out idea, and all of my professional experience leads me to be highly skeptical of the idea unless I see some more reliable evidence".
Extraordinary claims require extraordinary proof, as always.
BadBios could just as likely turn out to be an art piece.
Sadly the longer it goes without something the more difficult it is to believe. Time will tell of course, hell of a way to ruin a reputation though if its bogus.
I'm not doubting Dragos's awesomeness; I'm doubting how much bearing his awesomeness has on the truth value of the badbios narrative. The press getting wind of this and turning it into a sensational story didn't help.
Stuxnet was floating around the net for about a year before there was public discussion. It was only afterwards that people went back and realized that something had been there all along.
Part of the problem with expecting people to notice this class of viruses is that they are not malicious, don't use excessive amounts of resources and are deliberately designed to be stealthy. They aren't a general purpose bot-net, they don't target anyone's bank accounts -- in short they have the lowest possible profile that a well-funded and well-managed engineering organization can come up with.
From that perspective, the only odd sounding behavior is that infected systems avoid booting from anything but USB drives. That's a little high-profile for this class of virus. It isn't enough for me to dismiss Dragos's claims, but it is incongruous with the assumed design goals.
We must support researchers who behave like researchers. I wouldn't have made the claims he's made without first obtaining ... some evidence. I'd expect people to call me a liar or a loon if I made claims like these without a shred of evidence. This seems more like an art project than research.
Hook the USB to a logic analyizer, boot a disk in an in circuit emulator. Something. ... Don't just speculate about the sizes of ttf files in windows.
If it's a real threat, he's really hurting his case by ignoring the scientific principles outlines in this article.
Just because it's possible for someone to pick my lock, doesn't mean my belongings have been stolen. To conclude that my belongings have been stolen, we must first establish that they are actually missing, and that I did not lose them on my own. We do not immediately focus on the viability of lock picking. Everybody wrapped up in this story skipped this important step when examining the circumstances.
It turned out that when Dragos supplied what he thought was an "indicator of compromise" as it's called in the industry, or suspicious activity, in every instance it in fact was not suspicious at all, and Dragos either made a technical error or misunderstood what he was looking at.
You don't have to be a paranoid crazy to make this mistake, so the John Nash comparisons are over the top. I've been there in my career, which is partially why I recognize it in him. Unfortunately he has committed too strongly to this position, so he's not able to walk himself back from his early conclusions. I'm not at all naive or misinformed as to what malware capabilities could be (and I have some healthy paranoia myself), it's just that it's completely besides the point in this case.
"Urr...nothing, really". "Come on, you gotta know something. Is our hardware vulnerable?". "Urrr...dunno. We don't know if it exists, even". "But but, Ars Technica, ErrataRob and Schneier have written about it?" "Sigh, yes, I know."
This is why I've been vocal about getting these data confirmed. It's not about bashing Ruiu, it's not about being negative, it's not about whether I myself am able to verify everything - it's about trying - and failing - to get people to not lose their heads before the data are in.
So please don't frame this as "naysayer" discussion. If you shout fire in a crowded theater, you better have something to back it up with.
Remember: There are already documented cases of virii making themselves almost entirely undetectable, crossing air gaps on USB sticks, and then blowing up, e.g., specific brands of peripherals (such as industrial centrifuges) by altering their firmware.
Is it really a stretch to imagine a variant that (a) is bios-based and (b) uses the tech behind every modem since 300 baud -- the modulation of digital data into sound? That breaks itself up into undetectable pieces? We've seen these techniques before.
a. BIOS are all insanely different. A motherboard manufacturer can switch ethernet controllers within a run of a single board and ship an entirely new BIOS. You would need to generate thousands of viral stubs to get the infection spread being claimed, and the malware would need to move around GB of these BIOS so it has them handy when it runs into a new motherboard.
b. Researchers have worked on this and demonstrated it was possible. They had to modify off the shelf hardware to make it work. PC speakers are not designed to emit frequencies outside the audible range. It's like saying "we have the technology for cell phones, and my home router has a radio built in, it must be able to talk to cell towers!"
The situation where I see his claims being plausible is if a state actor has broken in and modified his hardware, and combined it with a viral threat custom built for what they know he has in his home.
I would like to read about this experiment you mention, do you have a link?
But my main point is - viruses (and other forms of malware) aren't actual biological viruses necessitating strict containment. They're code and we CAN understand them. More access means publicly accessible countermeasures are likely to develop faster. And yes - some people will take advantage of this by adapting the techniques employed therein for evil. I think it's worth the risk to try to understand it.
And yeah, don't just run it on your machine directly in a debugger of course. That's likely to be containment enough.