Re: Impending Crypto Monoculture
Andrew McGlashan via luv-mainwrites: > On 15/04/2016 4:51 PM, Rick Moen via luv-main wrote: >> Quoting Andrew McGlashan (andrew.mcglas...@affinityvision.com.au): >> >>> letsencrypt perhaps? It works very well. >> >> It (https://letsencrypt.org/, a recently invented, automated, >> no-charge >> CA) solves the one specific problem it set out to solve, well. And >> it's >> commendably well intended & benevolent. >> >> [But the CA model is incorrigibly broken.] https://en.wikipedia.org/wiki/Trust_on_first_use This model has worked well for OpenSSH for a long time. There is some recent(ish) discussion about applying it to "the web": https://blogs.fsfe.org/jens.lechtenboerger/2014/03/10/certificate-pinning-with-gnutls-in-the-mess-of-ssltls/ https://blogs.fsfe.org/jens.lechtenboerger/2014/03/23/certificate-pinning-for-gnu-emacs/ https://blogs.fsfe.org/jens.lechtenboerger/2014/04/05/certificate-pinning-for-gnulinux-and-android/ Short version is: it's not ready for "normal" users. > Still, I've used self-signed certs too over the years and only > occasionally tried out other options ... for me, right now, > letsencrypt > is better due to how the main browsers are setting up users to > distrust > anything that doesn't come from a CA (however untrustworthy CAs might > be). Making your own autonomous CA (and creating certs from it) is not much harder than making a self-signed cert. The GNUTLS manual essentially explains exactly how to do it, and it's CLI options are *vastly* clearer than the OpenSSL ones. ___ luv-main mailing list luv-main@luv.asn.au https://lists.luv.asn.au/cgi-bin/mailman/listinfo/luv-main
Re: Impending Crypto Monoculture
On 16/04/2016 3:22 AM, Robin Humble via luv-main wrote: > On Sat, Apr 16, 2016 at 02:35:20AM +1000, Andrew McGlashan via luv-main wrote: >> How about having fingerprints saved in DNS records, self-signed or >> "official" CA signed certs ? > > Certs for the domain in DNSSEC for the domain. sounds good. > I've heard folks who know a lot more than me about protocols and > security discuss it favourably. > is it an RFC? Good question, don't know the answer. But I've just thought of another solution. Every single domain has a registrar. Each registrar should be required to offer proof of ownership of the domain so that we can get certificates easily. You have a "proof" of ownership in the form of signed data only accessible from the login for the domain management. You sign this data with a GPG key for an email that is set up on one or more of the domain's contact names (every contact type should be able sign the request). The public key identifier for that email resides on record at the registrar (not the private key of course). Now, you don't need to farts around with stopping / starting servers or use manual processes to attain certs that require you to jump through lots of hoops. Your CSR can include your proof from the registrar (signed by them), then counter signed by you. Then Let's Encrypt can provide the required certificate(s) and chain files. Automation of letsencrypt can be problematic, you need too much extra rubbish installed for that. The manual process (for use via servers that don't host the website(s) or mail services for instance) is still labour intensive; having to create files in well known areas and populate them with specific data and repeat that every time you renew the certificate. Whilst we continue to have a CA system, then it may as well be made less painful than even letsencrypt has managed thus far. The latest thoughts also lessens the chance that anyone with access to a web server for a domain name may create certificates without proper authority. Perhaps the registrar can also be expected to keep certificate fingerprints available on line (for each and all certificates relating to the domain names), but this might be a step too far. More thoughts? Cheers AndrewM signature.asc Description: OpenPGP digital signature ___ luv-main mailing list luv-main@luv.asn.au https://lists.luv.asn.au/cgi-bin/mailman/listinfo/luv-main
Re: Impending Crypto Monoculture
On Sat, Apr 16, 2016 at 02:35:20AM +1000, Andrew McGlashan via luv-main wrote: >How about having fingerprints saved in DNS records, self-signed or >"official" CA signed certs ? Certs for the domain in DNSSEC for the domain. sounds good. I've heard folks who know a lot more than me about protocols and security discuss it favourably. is it an RFC? cheers, robin ___ luv-main mailing list luv-main@luv.asn.au https://lists.luv.asn.au/cgi-bin/mailman/listinfo/luv-main
Re: Impending Crypto Monoculture
On 15/04/2016 4:51 PM, Rick Moen via luv-main wrote: > Quoting Andrew McGlashan (andrew.mcglas...@affinityvision.com.au): > >> letsencrypt perhaps? It works very well. > > It (https://letsencrypt.org/, a recently invented, automated, no-charge > CA) solves the one specific problem it set out to solve, well. And it's > commendably well intended & benevolent. So many wise words, Marian flu or not. Still, I've used self-signed certs too over the years and only occasionally tried out other options for me, right now, letsencrypt is better due to how the main browsers are setting up users to distrust anything that doesn't come from a CA (however untrustworthy CAs might be). GPG signed certs, not likely; there are plenty of other considerations, but GPG signing /may/ be part of a greater solution. How about having fingerprints saved in DNS records, self-signed or "official" CA signed certs ? I wish trust in computers (not just CAs) wasn't so broken. :( Cheers AndrewM signature.asc Description: OpenPGP digital signature ___ luv-main mailing list luv-main@luv.asn.au https://lists.luv.asn.au/cgi-bin/mailman/listinfo/luv-main
Re: Impending Crypto Monoculture
Quoting Andrew McGlashan (andrew.mcglas...@affinityvision.com.au): > letsencrypt perhaps? It works very well. It (https://letsencrypt.org/, a recently invented, automated, no-charge CA) solves the one specific problem it set out to solve, well. And it's commendably well intended & benevolent. But, IMO, the entire CA model is unfixably broken, _so_ Let's Encrypt is a benign attempt to prop up a hopelessly bad CA framework that needs to just die. For details, rather than my recapping the conversation I had about Let's Encrypt just this past month, please see: http://linuxmafia.com/pipermail/conspire/2016-March/008389.html Further downthread discussion: http://linuxmafia.com/pipermail/conspire/2016-March/008390.html http://linuxmafia.com/pipermail/conspire/2016-March/008391.html http://linuxmafia.com/pipermail/conspire/2016-March/008392.html -- Cheers, "My life has a superb cast, Rick Moenbut I cannot figure out the plot." r...@linuxmafia.com -- Ashleigh Brilliant McQ! (4x80) ___ luv-main mailing list luv-main@luv.asn.au https://lists.luv.asn.au/cgi-bin/mailman/listinfo/luv-main
Re: Impending Crypto Monoculture
On 15/04/2016 4:15 PM, Rick Moen via luv-main wrote: > https://linuxmafia.com/~rick/faq/dan-brandishing-legal-threats letsencrypt perhaps? It works very well. Cheers A. signature.asc Description: OpenPGP digital signature ___ luv-main mailing list luv-main@luv.asn.au https://lists.luv.asn.au/cgi-bin/mailman/listinfo/luv-main
Re: Impending Crypto Monoculture
Quoting Chris Samuel (ch...@csamuel.org): > It's an interesting situation, though I think I'd trust Dan a bit more than I > trust the USG now. :-) I trust Dan a _great_ deal more than I do the USG, and that's after he sort-of-almost-threatened a bogus lawsuit against me 2001 for committing 'libel' [sic] against his software on my Web page (and acting 'against the law' on it), the Web page I'd pointed people to from 1999 onwards, rather than continually repeating why I had disliked adminstering qmail for a living during 1999 (back when qmail was still relevant). https://linuxmafia.com/~rick/faq/dan-brandishing-legal-threats That being said, Guttman's point is extremely well taken that we need more than just djbware. Personally, I'm all for being a patent scofflaw, if necessary. -- Cheers, "My opinions may have changed, Rick Moen but not the fact that I'm right." r...@linuxmafia.com -- Ashleigh Brilliant McQ! (4x80) ___ luv-main mailing list luv-main@luv.asn.au https://lists.luv.asn.au/cgi-bin/mailman/listinfo/luv-main
Re: Impending Crypto Monoculture
On Friday, 15 April 2016 9:46:40 AM AEST Chris Samuel via luv-main wrote: > There is an interesting rebuttal of the fanboyism remark in the LWN article > in one of the comments here: Grr... hit the shortcut to send once, got a popup dialog saying that this shortcut can lead to accidentally sending messages and giving me a range of options, the first of which was to disable the shortcut, which I did. Hit that again and it sent it without prompting! Sigh... Here's the link I was about to insert. https://lwn.net/Articles/682260/ cheers, Chris -- Chris Samuel : http://www.csamuel.org/ : Melbourne, VIC ___ luv-main mailing list luv-main@luv.asn.au https://lists.luv.asn.au/cgi-bin/mailman/listinfo/luv-main
Impending Crypto Monoculture
Hi folks, Given the ongoing crypto discussion I thought this post on LWN "The prospect of a crypto monoculture" from a few weeks back (publicly available now) might be of interest: https://lwn.net/Articles/681615/ It's a discussion of Peter Gutmann's post to comp.encryption.general entitled "On the Impending Crypto Monoculture" about how we are heading towards a situation where the only strong encryption that seems well designed is both designed and implemented by a single group, led by Dan Bernstein. There is an interesting rebuttal of the fanboyism remark in the LWN article in one of the comments here: Peter's post is archived here: https://lwn.net/Articles/681616/ Peter has been doing crypto for a long time and wrote the excellently titled "Everything you Never Wanted to Know about PKI but were Forced to Find Out": https://www.cs.auckland.ac.nz/~pgut001/pubs/pkitutorial.pdf It's an interesting situation, though I think I'd trust Dan a bit more than I trust the USG now. :-) All the best, Chris -- Chris Samuel : http://www.csamuel.org/ : Melbourne, VIC ___ luv-main mailing list luv-main@luv.asn.au https://lists.luv.asn.au/cgi-bin/mailman/listinfo/luv-main