oss-sec mailing list archives
Re: can we talk about secure time?
From: Daniel Kahn Gillmor <dkg () fifthhorseman net>
Date: Sat, 20 Dec 2014 11:33:22 -0500
On 12/20/2014 09:42 AM, Stuart Henderson wrote:
On 2014/12/20 12:27, Hanno Böck wrote:Is there any reason not to tell everyone to use tlsdate? What's the distro's take on this? afaik many ship ntp-based solutions by default.That won't work well for OpenBSD; libressl uses a random value instead of the timestamp.
It's not just libressl, there's a general push to do this: https://tools.ietf.org/html/draft-mathewson-no-gmtunixtime-00 and the upcoming TLS 1.3 is likely to have the timestamp removed from the handshake entirely: https://tools.ietf.org/html/draft-ietf-tls-tls13-03#section-1.2 tlsdate can also fetch the Date: header from https connections, though, so that should make it possible to withstand the removal from the handshake. Even if the time were to remain in the handshake (or we're ok with using the HTTP Date: header), tlsdate has a chicken-and-egg problem: if you want to actually verify the TLS connection, you need to validate a signature from a valid X.509 certificate; but knowing the validity of any given certificate is dependent on knowing the correct time. the default installation of tlsdate on debian jessie (0.0.12-1) appears to list 143 CAs in /etc/tlsdate/ca-roots/tlsdate-ca-roots.conf (all the usual suspects :/) so it doesn't appear to be using any targeted cryptographic keying material beyond the general public CA expectations. if we're going to solve the network time situation with cryptographic mechanisms, using time-bounded certificates for validity seems problematic. That said, for systems which need only rough precision and stronger network security (which is most systems), tlsdate does a much better job at the moment compared to ntp. (ntp offers more precision and less security, which seems like the wrong tradeoff for most internet-connected devices) --dkg
Attachment:
signature.asc
Description: OpenPGP digital signature
Current thread:
- can we talk about secure time? Hanno Böck (Dec 20)
- Re: can we talk about secure time? Stuart Henderson (Dec 20)
- Re: can we talk about secure time? Daniel Kahn Gillmor (Dec 20)
- Re: can we talk about secure time? ncl () cock li (Dec 20)
- Re: can we talk about secure time? Daniel Micay (Dec 20)
- Re: can we talk about secure time? Florian Weimer (Dec 21)
- Re: can we talk about secure time? Daniel Micay (Dec 21)
- Re: can we talk about secure time? Dave Horsfall (Dec 21)
- leap seconds and security [was: Re: can we talk about secure time?] Daniel Kahn Gillmor (Dec 21)
- Re: can we talk about secure time? Florian Weimer (Dec 21)
- Re: can we talk about secure time? Hanno Böck (Dec 21)
- Re: can we talk about secure time? Kurt Seifried (Dec 21)
- Re: can we talk about secure time? Hanno Böck (Dec 21)
- Re: can we talk about secure time? Stuart Henderson (Dec 20)