25C3: Dan Kaminsky Invokes DNSSEC

Kaminsky's talk at the Berlin hackers' meeting, that by its 25th year has become something of a legend, crowded an audience into the seats and aisles of Berlin's large Congress Center that could rival that of a keynote at an applied economics conference during the boom years. Without going into muchdetail on the DNS vulnerability issue per se (such as we have covered here), Kaminsky nevertheless announced some work of his that was under way. With the help of Internet veteran Paul Vixie, he had assembled around 15 DNS experts from the leading software makers into a room at Microsoft to discuss the consequences. The important thing for the participants was the simplicity of the solution so as to gain a better acceptance among system admins. If the DNS semantics would change even enough for just a percentage of hostnames not to be resolved, admins wouldn't bother to institute a patch. They would typically wait a few weeks after a vulnerability were announced before patching maybe half their systems. At that rate, about 75% of the systems would be patched, said Kaminsky proudly.

In retrospect, Kaminsky considers 2008 "not a good year for authentication." There were too many security holes whereby systems insufficiently tested the validity of domain requests. DNS was a big part of the problem, directly and indirectly, because many other Internet services, such as email and the WWW, depend on DNS. Many more applications participate in DNS Federations, which decentralize the DNS data over multiple authorities. Kaminsky provided a few examples: IPSEC keys, antispam blacklists and VoIP phone numbers. DNS was not originally designed for these solutions.

Kaminsky, who covered his topics more as an entertainer at times, didn't have any specific patent initiatives in mind. DNSSEC mightn't be elegant, but after all, could be part of the solution, by his opinion. Daniel J. Bernstein, well known in developer circles for his flawless code, but also feared for his release politics, presented his own solution, known as DNSCurve. Kaminsky nevertheless still saw a problem with the concept, because every request would then require too extensive a cryptographic operation. He doubted that heavily frequented nameservers could handle the load. It's yet to be seen whether in the foreseeable future we will truly encounter the oft repeated "end of the Internet as we know it."