Experts scramble to quash IPv6 flaw
'Can be exploited by any greedy Estonian teenager with a $300 Linux machine'
"IPv6 is really neat, but I think we are going to see a number of these gotchas because it is still so new," said Jose Nazario, senior security researcher with Arbor Networks. "It will likely shake out over the next couple of years."
Under IPv6, the impact of allowing users to specify some of the addresses to which data must be sent, known as loose source routing, is more dire. Because more addresses can be included in the header, rather than magnifying an attack by 10, Biondi and Ebalard calculated that it could amplify attacks by a factor of 88. In addition, RH0 also could allow an attacker to dodge a distributed technology, known as AnyCast, for protecting the 13 DNS root servers from attack and could be used to create a backlog of packets that could spike traffic to a server at a specific time.
"It is exactly that: The reintroduction of the IPv4 loose source routing mechanism in the IPv6 world and on steroids," said a network engineer who asked not to be identified.
The IETF reaction may have set a new speed record for the standards-setting body. With engineers arguing technical merits and peer-reviewing others' work while vendors push their specific requirements, the IETF is not known for making quick decisions.
"In practice it, it will be disabled, whether it gets left around for future usage, that's up in the air," said Robert Hinden, co-chair of the IPv6 Working Group for the IETF and a Nokia Fellow at the networking and phone giant.
Yet, companies and the engineering group responsible for a large portion of the IPv6 routing code have moved quickly to disable the feature. By late April, the Kame Project, which has created the code used in many flavors of the BSD operating system as well as routers, had disabled the Type 0 Routing Header in its own code.
"They don't just avoid walking the RH0 header, but they also now drop packets that contain it," said Theo de Raadt, project leader for OpenBSD.
Cisco has issued a security advisory on the issue. Both Cisco and Juniper declined to provide a representative to discuss the issue.
Because IPv6 has not been fully deployed in most networks, it will likely only take two or three years for almost all Internet service providers to fix the issues, de Raadt said.
ISC's Vixie agreed that the problem should be almost completely eliminated in three years.
"I'd say in three years this will be a footnote," Vixie said.
This article originally appeared in Security Focus.
Copyright © 2007, SecurityFocus