Kaminsky, who discovered the now-infamous big DNS flaw last year and got the vendor community to patch it, had for some time mostly dismissed DNSSEC as a DNS security solution. But after studying the specification more closely, Kaminsky -- who discussed his newfound support for DNSSEC here during his Black Hat DC presentation -- said DNSSEC could remedy some of DNS' security weaknesses.
"I was never anti-DNSSEC -- I was just never for it. It just didn't look like it was going to work," Kaminsky said in an interview today. "This is my first time publicly saying we need to do it. No one is more surprised than I."
The federal government is already on its way to widespread DNSSEC adoption after initially only recommending it for some systems. A new federal policy (PDF) issued in the wake of the DNS flaw scare last summer mandates that all federal agencies adopt DNSSEC by December 2009 for their DNS servers.
DNSSEC has been criticized for its complexity, as well as the DNS infrastructure overhaul its adoption typically entails. For DNSSEC's validation model to work for DNS servers, it has to be adopted from end to end, for instance, an issue that is rife with both technical and political challenges.
And DNSSEC requires significant manual configuration, including the signing of encryption keys and updating records, Kaminsky noted.
"The protocol isn't bad. It just has too many [manual] knobs," Kaminsky said. "We have to figure out how to shrink the number of knobs and to automate this. And this is not an impossible task."
DNS servers should be able to generate encryption keys and automatically sign records themselves, for example, he said. And DNSSEC also must be made more scalable than it is today.
"The implementations out there today are not fun to deploy," he said during his presentation. "We need to start migrating new apps to DNSSEC, so end to end you are able to use federated trust."
Why the change of heart for Kaminsky on DNSSEC? "These [DNS] attacks are no longer theoretical," he said. Somewhere between 1 and 3 percent of unpatched servers have already been hit by cache-poisoning attacks, according to data gathered recently by Georgia Tech data, and those are only the ones the researchers are monitoring. Meanwhile, one-third of all servers remain unpatched for the big DNS bug, he said.
It's time to get DNSSEC rolled out -- but with some key improvements, he said. "My argument is that we need to make DNSSEC deployable today," Kaminsky said.
Among those improvements, according to Kaminsky: The most popular type of DNS server, BIND, must be retrofitted to better support DNSSEC, and root servers need to digitally sign or authenticate DNS traffic. "We also need to work with the interaction among registrars," he said. "DNSSEC solves the problem of getting secure data back out, but how do you get data into .com and .org? Authoritative [DNS] servers today don't interact with their parents...Name servers need a way to automatically inform their parents [.com, .org] that they have the keying material."
Even with these issues, DNSSEC is still the best solution for cross-organizational authentication. "DNSSEC is the key to fixing the persistent authentication problems plaguing real-world, cross-organizational business for years," Kaminsky said.
Have a comment on this story? Please click "Discuss" below. If you'd like to contact Dark Reading's editors directly, send us a message