Kaminsky (finally) reveals gaping hole in internet
DNS patch averts doomsday scenario
Black Hat After a four-week orgy of speculation, recrimination and warnings, Dan Kaminsky's domain-name system vulnerability has finally gone public. And boy, are we glad the net's overlords paid attention.
During an 80-minute presentation, Kaminsky for the first time gave a detailed analysis of a bug that threatened to bring chaos to the internet by poisoning the machines that translate domain names into internet protocol addresses. The director of penetration testing for security firm IOActive first warned of the threat on July 8, but withheld specifics out of concern that they would make it easier for miscreants to carry out attacks.
As other security researchers have publicly speculated, the DNS vulnerability stems from shortcomings in the way servers try to ensure IP address information comes from bona fide sources rather than those controlled by miscreants. To prevent tampering, DNS queries include a random transaction number. The response is only considered valid only if it contains the same number.
What Kaminsky discovered is that this safeguard can be easily circumvented because there are only 65,536 possible transaction IDs. By flooding a DNS server with multiple requests for a domain name with slight variations - for instance, 1.google.com, 2.google.com, 3.google.com and so on - an attacker can vastly improve the chances of reproducing the correct number.
"Because a bad guy can repeatedly cause these races to the same name server, he's eventually going to win," Kaminsky told a standing-room-only crowd.
Whereas the chance of guessing the correct transaction ID from a single query is 1 in 65,565, the odds improve considerably when an attacker floods a server with thousands of requests in a short burst.
The details were enough to satisfy us, and plenty of Black Hat attendees, that the past four weeks of handwringing was warranted.
"The fuss was justified from the perspective that this is an impactful finding that has the potential to bring down the internet," said Nitesh Dhanjani, a senior manager at Ernst & Young.
While many organizations have heeded Kaminsky's warning to patch, a disturbing number have yet to fix their servers. Among Fortune 500 companies, 15 per cent have yet to take any action, and another 15 per cent are still vulnerable to some extent because they use network address translation gear that prevents the patch from working.
Wearing his trademark blue jeans and black shirt and sneakers, Kaminsky laid out a plethora of ways miscreants could exploit the bug. Hitting mail exchange servers, it could be used to intercept or tamper with large organizations' email. It could be used alongside other attacks such as sidejacking and the manipulation of auto-update features in third-party applications.
"There are a ton of different paths that lead to doom," Kaminsky said.
While there has been only one confirmed attack using the flaw, Kaminsky suggested there likely were others.
"There's a lot of weird stuff floating around out there," Kaminsky said, referring to recent analysis of DNS traffic.
In the five months since he discovered the flaw, Kaminsky has shouldered considerable burdens in trying to get it addressed. He spent countless hours trying to marshal engineers from Microsoft, Sun Micro, and dozens of other companies. And he's endured criticism that he shamefully exaggerated the threat in a cynical attempt to drum up hype for his Black Hat presentation.
Given the huge sums being paid for unpatched vulnerability disclosures to widely used systems, we'd hate to think what might have happened if a less scrupulous person had stumbled on the bug first, or for that matter whether Kaminsky would be willing to undergo the same trials the next time he discovers a flaw of this magnitude.
"I might be," Kaminsky said. "I don't know if my girlfriend is. She's been giving me the please-don't-break-the-internet-again look." ®
Sponsored: Benefits from the lessons learned in HPC