The dirty little secret about security bugs is not every single vulnerability that gets reported is exploitable -- meaning there are some that an attacker can't use against you. So how do you know which threats to patch right away and which to ignore?
There's no real formula for this, and even some security experts are divided on which bugs are not actually exploitable. "There is a lot of controversy among researchers about which announcements really pan out to exploitable vulnerabilities," says Thomas Ptacek, a researcher with Matasano Security. "When researchers publish advisories, people suspect the flaws are harder to exploit than they sound. When vendors publish, people suspect it's easier. The balance probably favors researchers.
"I'd guess 80 percent of what's announced turns out to be readily exploitable, and 10 to 15 percent more are refined into something dangerous down the road."
Microsoft's Security Response Center (MSRC) vets each and every bug that gets reported for its products. "We'll investigate any reported vulnerability sent to us, from a third-party ISV, a hacker mailing list, whatever," says Mike Reavey, a security program manager at the MSRC.
Those vulnerabilities Microsoft patches and considers the most dangerous are labeled as "critical," and ones that are harder to exploit get a lower severity rating, he says.
But sometimes, bugs that get reported just don't pan out, says Mark Griesi, a security program manager at the MSRC. "There are times when things that are reported as vulnerabilities quite frankly are just not," Griesi says. "Someone might report something that can circumvent a security measure in the OS, for example, but in reality, it requires someone sitting at the machine doing it themselves, on purpose. That isn't a vulnerability."
Some researchers are skeptical about the weaker or non-exploitable flaws: David Maynor, CTO of Errata Security, says not all bugs require you to buy a new security tool or to drop everything and patch. "Just because you find a bug doesn't mean it's going to be exploitable," Maynor says. "That's most of what our [Errata's] business is about: Do you really have to stop and buy new security to fix this... if [an attacker] can't do anything useful with it?"
"Not all software flaws are security vulnerabilities," he says. Just because an attacker can take advantage of a weakness or hole doesn't mean he can do any damage with it, Maynor notes. He points to the recent Universal Plug and Play bug in Microsoft's XP. Errata found in its research that the bug wasn't truly exploitable, he says. Maynor considers Microsoft's recent DNS Server Service bug -- also "critical" -- as serious, however. "The DNS bug was a big deal."
RSnake, a.k.a. Robert Hansen, founder of SecTheory, says sometimes you just can't get at a bug. "I remember one researcher telling me he's just waiting for someone to 'fix' an issue which will make the browser exploitable."
Worthless bugs don't get reported because they'd get laughed out of the security community, anyway, he says. "Most people aren't interested unless it's [the bug] remotely exploitable... That's pretty short-sighted, but it is what it is."
Microsoft's Reavey notes that not every critical vulnerability Microsoft patches will actually get exploited by an attacker.
But Microsoft doesn't try to outguess whether an attacker will or will not go for it. "Just because they haven't figured out how to exploit it today doesn't mean they won't figure it out tomorrow or a year from now," Microsoft's Griesi says. Microsoft doesn't take any chances with potentially critical bugs, he adds.
Researchers agree that in the end, it's better safe than sorry: "Even the most dedicated researchers can't quickly tell you whether a new memory corruption finding is absolutely not exploitable -- there's no silver bullet answer," Ptacek says. "Enterprises should probably assume the worst about every advisory."Says Maynor: "It's never a bad thing to install patches."
Kelly Jackson Higgins, Senior Editor, Dark Reading