On Tue, Feb 1, 2011 at 2:30 AM, Paul Wouters <[email protected]> wrote:
> On Tue, 1 Feb 2011, Brian Dickson wrote: > > This may be good enough for DNSSEC purposes. >> > > At least to then do ntp and and see that it matches our rough expectation. > Though in all, if the attacker is your controlling upstream, you are lost. > > Paul Active man in the middle is rather easier than factoring a 2048 bit RSA. With active man in the middle it is going to be possible to perform a replay attack. Checking for consistency of data is going to make a replay attack harder but not impossible. This may not matter if the device that ultimately relies on the data returned has a trustworthy time source. Otherwise you need a mechanism that provides trusted time that is secure against a replay attack. That is at minimum going to require the ability to persist a trusted time response to provide a weak form of protection but strong protection requires a challenge response mechanism and trusted time. Since ICANN does not provide trusted time, an additional trust root is going to be necessary. One issue that does come to mind is that if people are going to implement the trustworthy resolver model, it would probably be useful to be able to use the DNS resolver as a low fidelity source of trustworthy time. -- Website: http://hallambaker.com/
_______________________________________________ DNSOP mailing list [email protected] https://www.ietf.org/mailman/listinfo/dnsop
