Comments on: Cryptographic Nonsenseware

By: Ivan Krstić

Ivan Krstić — Thu, 26 Jun 2008 16:07:10 +0000

Mike — I’ve talked to some of the top Windows security people about it, although not at great length. The consensus seemed to be that large parts could be ported with fairly low compatibility impact.

By: Mike Hearn

Mike Hearn — Mon, 23 Jun 2008 16:55:49 +0000

How much research has been done into retrofitting BitFrost like security models onto Windows? I know that CoreForce has done a decent job of bolting a kind of MAC security onto it, but what about a more complete solution?

By: martin langhoff

martin langhoff — Sun, 15 Jun 2008 23:46:14 +0000

“The OS vendors need to start thinking very hard about this, because the current defenses just won’t cut it.”

And we hope they don’t, because then we are out of a job ;-)

By: Ralph

Ralph — Sun, 15 Jun 2008 21:41:05 +0000

Ivan, I read your slides. Very good!

I really don’t think security has to be quite so difficult as it seems right now. The problem, as I see it, is not about figuring out what to lock and how to lock it. It’s about figuring out where to install a few carefully-placed doors into otherwise solid walls.

I used to write a lot of code destined for ROM. I believe ROM code can be made intrinsically safe against any attacker who does not have physical access to our hardware. (One exception is for content-free denial of service attacks, which cannot be prevented by any locally-deployed method.)

Of course no system can ever be completely secure. As users, we will always be able to act carelessly and take imprudent risks. Nevertheless, we can do better than to leave all our money and possessions out on the front porch every night, with fireflies assigned to guard them — which is essentially what we’re doing now.

By: Ivan Krstić

Ivan Krstić — Sun, 15 Jun 2008 21:31:45 +0000

Martin — clearly it’s not the place of an anti-virus company to claim they’ll develop a better OS. But their PR is just disingenuous, and I fault them for it — for not saying “Look, we got lucky this time because the key length is such that we can conceivably factor it, but next time, we won’t get lucky. And we don’t have a solution to that problem. The OS vendors need to start thinking very hard about this, because the current defenses just won’t cut it.”

By: martin langhoff

martin langhoff — Sun, 15 Jun 2008 21:26:23 +0000

I agree that it is lame (but useful to some extent) to attempt to factor the key. Customers affected by the current run may be able to get their files out. In some cases a completely lost file, recovered in a few months, is better than no file.

However… if Kaspersky’s PR had said “ah, we will take a few years to write a new OS, based on safe paradigms; users who migrate to it will be safe from this threat”, now that would be elegant useless crystal tower talk. They aren’t a OS outfit, like MS, RH, Debian/Ubuntu, Novell or OLPC. And they can’t turn their backs on their customers like that.

Frankly, I am not sure if there is something practical that can be done in the current situation to help people _now_. Those users are in the sh*t, and will continue to be in there for a while — yes, hopefully the industry will move faster towards a better paradigm… but however fast it happens, it won’t be fast enough :-/

By: Ivan Krstić

Ivan Krstić — Sun, 15 Jun 2008 17:10:39 +0000

Ralph — there are no better countermeasures from the anti-virus software point of view. The improvement has to come from the people who make operating systems; Bitfrost, the OLPC security architecture I designed, is one such effort. You might find my AusCERT 2007 slide deck useful to clarify things.

By: Ralph

Ralph — Sun, 15 Jun 2008 17:05:19 +0000

Ivan — I do think the antivirus companies should get into the full-disk backup business — or whatever kind of backup is required at this point. That’s only sensible.

With respect to “there are no better countermeasures,” didn’t you just write that you are working on such a system for OLPC?

Personally I just cannot believe that an OS is impossible tod secure. I mean, suppose all the code runs from ROM, and it is (physically) impossible to execute anything from data space. I don’t see how that arrangement could fail.

By: Ivan Krstić

Ivan Krstić — Sun, 15 Jun 2008 16:18:40 +0000

Sam — I think Symantec’s classification speaks to a different kind of containment and removal ease. When the malware finishes its encryption run, it’ll destroy itself leaving nothing but the ransom instructions behind, meaning it’s both self-contained and self-removed. This fails to capture the severity of losing all your documents, however.

Ralph — I considered the “we’ll help people already affected while we develop better countermeasures” argument, but the point is that there are no better countermeasures for anti-virus companies to develop. Unless they want to get into the full-disk backup business.

By: Ralph

Ralph — Sun, 15 Jun 2008 14:58:26 +0000

Of course you’re absolutely right about factoring being a dumb idea…

except that one could justify factoring a key which has already been imposed on some victims, while simultaneously taking entirely different and stronger measures to prevent this happening to anyone in the future.

The first and most logical method of rendering harmless future exploits of this type would be to make backups… something the victims should have been doing every day, ever hour or every minute, depending on how much currency they can afford to lose.