Slashdot Mirror
MS Giving Exploit Writers Clues To Flaws
In the IT trench writes "How's this for a new twist on the old responsible disclosure debate? Hackers are using clues from Microsoft's pre-patch security advisories to create and publish proof-of-concept exploits. The latest zero-day flaw in the Windows DNS Server RPC interface implementation is a perfect example of the tug-o-war within the Microsoft Security Response Center about how much information should be included in the pre-patch advisory."
Umm, do these people really think that hackers can't reverse engineer a patch and see what it's doing??
.. because they will have no idea what the patch is doing and if it will affect other stuff they have. Also, this will be an excuse to further delay the release of patches, since now it will have to go through even more QA.
.. now that's funny.
IT admins will be the most affected
Hackers that RTFM
It would have taken an open source project longer to write the security advisory than to change 1 line of code (or 2 tops) to fix the stack overflow issue.
How hard is it for Microsoft to push out a new update for a change this minor (and important)?
This is a critical problem for any intranet (Universities come to mind as the largest target) running Microsoft servers. And it can also affect a whole load of dedicated servers running the basic versions of Microsoft server software.
What are they waiting for?!
I seem to remember that one of the reasons XP was cracked was that the Windows XP team responsible for it gave an interview about it. They were so proud of it that they gushed about the nuances of it. Now, they didn't exactly give a lot to technical detail but they gave enough to crackers to figure out how it worked. Pretty soon there was a key generator program released.
Well, there's spam egg sausage and spam, that's not got much spam in it.
>>But this is a case where a half-and-half approach is probably the worst of all.
Of course, going halfway on either ideology, closed or open source sets you up for trouble. Which is why it is a bad idea for Microsoft to release any details about their patches at all. Yes, I said it. If they are going to try and make a blackbox OS, they shouldn't do it half way.
It tells me, as an administrator, what to be suspicious of, rather than some secret bug they are patching.
Now, it's true that it is still in the favor of the virus writers in that hardly 100% of sys admins keep up to date on this stuff (wheres the time?) but it is scary that they can exploit a specific bug base on a vague explanation in the first place..... (scary in that Windows is really that bad...)
Microsoft should pre-publish a whole bunch of tasty looking security advisories that are 100% fake every time they publish one that is real. Make them the most enticing looking (remote code exploit with unvalidated input overflow in ssh). Any given cracker will probably pick the fake and quickly waste gobs of time.
If they wanted to get more diabolical, they could even put some honey pots into the code itself. For example, something that emulates a buffer overflow crash when a certain malfromed word is injected. Or maybe something more tantilizing but useless like a 1 second pause in Internet explorer when a certain tag combination appears followed by a page reload to make them think IE just belched but managed to somehow recover. Hint at this in the pre-pub or leak it on the web (post it in a slashdot comment). they can validate it's existence so they believe the bug really exists too.
Each time they patch the real security hole they can preload ten new honeypots for the next round of spoofing the hackers and eradicate the old ones so it looks like they are patching real bugs and the hackers never catch on.
Why am I posting this under this parent? Well because you could only get away with this in closed source. Open source would make this a give-away.
Some drink at the fountain of knowledge. Others just gargle.
Parent is right - the DNS RPC flaw is a 0day issue discovered in the wild. It was reported to Microsoft, apparently by TippingPoint and the Zero-Day Initiative. That is why they released an Advisory (something they usually only do for issues discovered in the wild).
Put another way - it was actually initially discovered by the black hats, and an exploitation tool released and used, not confidentially reported to Microsoft under the "Responsible Disclosure" programme, or even publically posted to somewhere like Full-Disclosure.
The article is talking obvious stuff. Obviously lame stuff. Just saying "There's a flaw in the RPC management protocol of Windows' DNS server" is enough to clue you in, because the bug is sitting in about the first place you'd look, a nice juicy classic stack overflow. That's also absolutely critical information that is required to effectively mitigate the attacks in the wild before they get any worse, and mitigation shouldn't normally be a problem (RPC management of DNS isn't that widespread).
As a black hat, guessing details of upcoming patches by looking at the pre-patch advisories is a poor idea:
1. They'll be patched, and not even a week later if you're looking at the regular notifications. That's a pretty limited timeframe to exploit the bug.
2. Even worse, MS can push it out early if they really have to, because generally by that time the patch has already been completed for a month or so and is, in fact, sitting in testing (go ahead, check the dates on the digital signatures if you don't believe me). And the development/fixing QFE branch is quite a bit ahead of the stable GDR branch. Sometimes you find that by analysing the differences between them, interesting changes pop up.
You're better off looking at past exploits and trying variations on them, because MS have a habit of incompletely fixing a bug by fixing one code branch, but missing another closely related one on the very same page. For example, patching a length buffer overflow in the first ANIH chunk of a RIFF file, but forgetting they have another parser (that hasn't been patched) for handling subsequent ANIH chunks. (If this sounds familiar now, it should.)
You're even better off looking for novel exploits in the old-fashioned way; target any code that parses or handles data in any form that you can manipulate, figure out which of the potential targets represents the juiciest opportunity for a vector, and start either disassembling and looking for oversights, or blindly fuzzing boundary conditions and trying to make it break.
In that way you can stumble upon entirely original high-value exploits that bear little similarity to an existing issue and so are unlikely to be picked up by AV signatures or IDS until after the exploit is fairly well known or unless it hits one of the larger honeypots like dshield and they figure out what's going on.
Let us take a look at the recent topic of a Madwifi vulnerability affecting certain wifi users in Linux.
Julien Tinnes reported it at 13:48:00 EST on December 7, 2006.
At 14:17:50 on the same day the patch was available in the main source code repository.
A little while later at 17:08:26 the vulnerability is officially confirmed by Madwifi and advisories had been prepared.
Looking downstream, the response times for an official fixes/advisories by distribution specific security teams were:
Gentoo: December 10
SUSE: Confirmed December 8, Fixed December 11
Ubuntu: January 9
There is certainly some room for improvement here with distribution specific fixes, but that also includes time spent testing the changes to the driver. To be fair to Microsoft (actually, I'm just being overly optimistic), they probably had a patch ready within 30 minutes of the initial vulnerability report as was the case with Madwifi. But instead of giving the customer the option of trying the "beta" patch so they can test it themselves, it is kept private. Days tick by at Microsoft HQ and nothing appears to happen. Eventually, a patch is released on the patch Tuesday of the next month (or the month after that). System administrators get no choice and no chance to test it themselves.