MINNEAPOLIS -- The Internet engineering community is grappling with what to do about a serious flaw in the DNS discovered this summer, and the ongoing debate brings to mind a famous quotation from Voltaire: "The perfect is the enemy of the good."
At issue is whether the group should use its resources to encourage DNS registries, ISPs and enterprises to upgrade to the ultimate DNS security solution known as DNSSEC; or whether it should tweak the DNS protocols to address the so-called Kaminsky bug as an interim step. The issue is being debated at a meeting of the IETF, the Internet's leading standards body, being held here this week.
In July, security researcher Dan Kaminsky discovered a DNS bug that allows for cache poisoning attacks, where a hacker redirects traffic from a legitimate Web site to a fake one without the user knowing. With DNSSEC, the IETF already has a solution to the Kaminsky problem and other known DNS vulnerabilities. However, DNSSEC hasn't been widely deployed, although it has been under development for more than a decade.
DNSSEC prevents hackers from hijacking Web traffic and redirecting it to bogus sites. The Internet standard prevents spoofing attacks by allowing Web sites to verify their domain names and corresponding IP addresses using digital signatures and public-key encryption.
The problem is that DNSSEC prevents Kaminsky attacks only when it is fully deployed across the Internet -- from the DNS root zone at the top of the DNS heirarchy down to individual top-level domains, such as .com and .net. Until then, Web sites remain vulnerable to Kaminsky-style attacks.
That's why some IETF participants are urging immediate action to address the Kaminsky bug, while others are hoping to use the publicity surrounding the discovery of the Kaminsky bug to promote DNSSEC deployment. "The open question is whether there are other measures we can take as operators of the DNS to improve forgery resilience, or are there changes to the DNS protocols that we should be making that are an interim step that aren't all the way to DNSSEC," explains Andrew Sullivan, co-chair of the IETF's DNS Extensions working group, which is discussing the matter. The working group is split on which direction to take. "We can't tell yet which way it will go," says Olafur Gudmundsson, the other co-chair of the group.
In recent weeks, IETF participants have submitted five documents to the DNS Extensions working group with proposed changes to DNS that would prevent Kaminsky-style attacks. "We've been trying to condense the proposals down to the working group to show what each of the changes would be, how they would help the situation, what the operational costs would be and what would break as a result of the changes," Gudmundsson says. "It's too early to tell if the group is going to coalesce around one of these proposals."
One option is for the IETF to do nothing about the Kaminsky bug. Some participants at the DNS Extensions working group meeting this week referred to all of the proposals as a "hack" of the DNS and argued against spending time and energy developing one of them into an Internet standard because it could delay DNSSEC deployment.
Other participants said it is irresponsible for the IETF to do nothing about the Kaminsky bug because large sections of the DNS will never deploy DNSSEC. "We can do the hack and it might work in the short term, but when DNSSEC gets widely used, we'll still be stuck with the hack," said IETF participant Scott Rose, a DNSSEC expert with the U.S. National Institute for Standards and Technology (NIST). "Personally, I'd like to see DNSSEC deployed because I think it's the best solution. But there are going to be places on the Internet that aren't going to do DNSSEC, and I think maybe we should look at an interim solution for [them.]"
Several IETF participants said the threat of Kaminsky-style attacks is real. A representative from Comcast said the ISP has seen "large numbers of cache poisoning attacks" attempted since August. NIST also has seen hackers try to exploit the Kaminsky bug. "People are trying the Kaminsky attack. They're trying to find recursive servers that can be poisoned. We think they are trying to get a list of these servers and sell them," Rose said.
IETF participants pointed out that DNS software packages from BIND, Nominum, Microsoft and NLnet Labs have added patches for the Kaminsky bug, and 75% of DNS servers have been upgraded to thwart Kaminsky-style attacks. The IETF also is putting the finishing touches on a best-practices document that outlines ways for DNS server operators to protect against spoofing attacks like those that exploit the Kaminsky bug.
The co-chairs of the DNS Extensions working group said they hope to make a decision on whether to change the DNS protocols in light of the Kaminsky bug before the group's next meeting, which will be held in San Francisco in March. "There's been an awful lot of urgency to the matter. We want to avoid creating a long-term problem that is caused by a hasty decision," Sullivan said. "There are big reasons to be careful here. The DNS is a really old protocol and it is fundamental to the Internet. We're not talking about patching software. We're talking about patching a protocol. We want to make sure that whatever we do doesn't break the Internet."
In related news, the U.S. federal government is making progress on its efforts to deploy DNSSEC. The Office of Management and Budget issued a mandate in August that requires all federal agencies to support DNSSEC. That order states that .gov must be cryptographically signed at the top level by January 2009; and that all subdomains under .gov, such as www.irs.gov, must be signed by December 2009. IETF participants said .gov already is being signed at the top level ahead of the deadline.
The National Telecommunications and Information Administration set a deadline of next week for receiving comments from industry and academia about how best to roll out DNSSEC on the DNS root zone. The IETF’s sister organization -- the Internet Architecture Board -- submitted a comment this week to the NTIA offering suggestions for speedy and successful deployment of DNSSEC on the root servers.
Getting the root signed is the "800-pound gorilla in the middle of the room," says IETF participant Paul Hoffman, an Internet security expert who sent a comment to the NTIA "Let's say the root is signed tomorrow. Let's say all the important top-level domains are signed. It's still no good unless all of the domains are signed. You can't just deploy DNSSEC. You have to deploy it universally."