Last Call Review of draft-ietf-dnsop-negative-trust-anchors-10
review-ietf-dnsop-negative-trust-anchors-10-secdir-lc-sheffer-2015-06-25-00

I have reviewed this document as part of the security directorate's
ongoing effort to review all IETF documents being processed by the IESG.
These comments were written primarily for the benefit of the security
area directors.  Document editors and WG chairs should treat these
comments just like any other last call comments.



This document defines a process where an ISP can declare that certain 


domain names (whose DNSSEC records are deemed misconfigured) are not 


covered by DNSSEC. The server then provides DNS resolution in response 


to client queries, where otherwise the server would have failed those 


queries.




Summary



This is more of a rant than a review, however it presents a security 


perspective that seems to be at odds with the operational-first 


perspective of the document.




Details



I am hearing that this is a controversial draft, and I can see why. The 


draft explains very well what motivates the proposed mechanism, and the 


motivation makes sense, especially if you are an ISP. But I think the 


draft gives excellent instructions on improving the security of an 


inherently problematic usage model.






As an individual or even as a company, the local ISP is not my friend. 


State-controlled ISPs in less developed countries have been known to 


steal traffic and fake certificates in order to attack their own 


subscribers. Commercial ISPs in more developed countries throttle or 


block certain types of traffic for various reasons. Moreover, the local 


ISP is best positioned to identify the owner of individual IP endpoints 


and perform point attacks on local subscribers. DNS is an obvious attack 


vector.






Even assuming that 99.9% of us will continue to trust ISPs for DNS 


resolution, IMHO the proposed solution would be better off with more 


automation and less celebration of "trained technical personnel". For 


example, the document has a SHOULD level requirement to test the NTA 


"periodically" in order to eventually remove it. How about an 


alternative that shifts the responsibility to DevOps or to the actual 


vendors, and also empowers the .1% who maintain their own resolvers:






"NTA implementors MUST attempt to validate the domain in question once 


every MINUTE for the period of time that the Negative Trust Anchor is in 


place, until such validation is again successful, and MUST remove the 


NTA as soon as that happens."






Similarly, I would guess the process of establishing an NTA could be 


automated, e.g. by querying multiple major DNS operators over a period 


of time (maybe operators that are known to run the same brand of 


resolver). In general, automating the process is more likely to 


encourage deployment of DNSSEC by smaller ISPs than adding complex 


manual "best practices".




Thanks,
	Yaron