https://news.ycombinator.com/item?id=6990602
https://news.ycombinator.com/item?id=5990288
Edit: this is a nice walkthrough of a simple RNG bug https://nakedsecurity.sophos.com/2013/07/09/anatomy-of-a-pse...
I think the article[ed:1] misses the point, btw. Yes, managing keys might be tricky - but it's not really rocket science. The thing that's hard is managing trust -- which key one trusts etc. The CA system for web is completely broken. I had some hope for cacert.org -- I think that model (perhaps expanded to include recommending signing gpg-keys as well as x509 certs) has a lot of merit.
I think web of trust is the only thing that can work for managing trust. But it needs to be accessible. Have post offices and banks sign gpg keys on when people come in with valid id. Cacert is a different take -- I like to look on it as a "strucutred eternal keysigning party". I trust that model a lot more than the classic CA model. But as it is based on the CA model, it suffers the same problems with centralized trust. Centralized trust is great for organizations, it's not so great for individuals.
I think the best model would be a world-wide web of trust for gpg, helped by formal and informal signing organizations (ie: like cacert, signinparties -- and with the help of banks, governments, DMV and similar institutions that traditionally help with issuing IDs). Then there should be support for anchoring DNS/CAs (and CAs for openssh) with gpg. So that if you trust someone is a representative of an organization linked to a domain name, you can trust them to autorize a CA for that name (there's technical details here, but I think the idea should be clear enough).
CAs go away, everyone can sign their own certs -- and there's an easy way to link x509 and gpg trust.
People will still lose their keys, and get invalid keys signed etc -- key management is hard. But the really confounding thing is trust -- and knowing how to determine which keys are "proper" keys for a given entity. That's really trust management, not (just) key management.
[ed:1 whops, that was the other article on making key management easy ;-) But I suppose this comment is relevant wrt how to make encryption more readily available...]
Almost nobody can use PGP/GnuPG properly and OTR is reasonably easy to use for most people. Isn't this alone a good reason to just tell people to ignore PGP/GnuPG altogether?
to whit:
Wed Jan 11 00:33:40 EST 2006, CLAY SHENTRUP
CLAY at BROKENLADDER.COM wrote:
>
> On 1/10/06, Daniel Guido <dguido at
> gmail.com> wrote:
>>
>> Correct me if I'm wrong, but there is no
>> working implementation of OTR for e-mail
>> yet is there?
>
> There can't be really. The sender and
> receiver have to agree upon a shared secret
> at the time of transmition, which requires
> at least 3 passes.So, again, what's wrong with telling people to ignore PGP/GnuPG altogether?
-- Assange to Google's Schmidt: 'I don't use email'
http://www.computerworld.com/article/2496908/encryption/assa...
And as Assange said, if you are e.g. a journalist in the government watchlist, it could be worse (more dangerous) than not using gpg/pgp.
Text messages using OTR seems like a better way. For widespread crypto usage, telling people to ignore pgp/gpg and use texts with OTR seems more reasonable than keeping projects like GnuPG alive.
If I were running a company like Facebook who's interested in spying on people, I might even fund projects like GnuPG so that unrealistic geeks keep thinking this is a viable solution.
As for facebook, as long as they keep the XMPP access open and supported[1], at least they do support OTR. Unlike eg: google.
I don't really understand this "gpg is impossibly hard"-stance. Yes, security is hard. Why recommend OTR? Don't get me wrong, I love OTR -- but verifying OTR keys, and transporting identities across devices (eg: when getting a new phone) is pretty difficult too. Are you saying wrong use of OTR is better than wrong use of gpg, because most people that use OTR use it in a way that allow for MITM anyway?
But most importantly, the point you seem to ignore, it's not secure if you don't verify keys. I'd say most people don't verify keys with OTR -- hence they're not using it in a manner that's actually secure. I do agree that it's a lot better than people using Snapchat.
What about the rest of the scorecard?
They factored the JS into a Chrome application awhile back, and while Chrome-extension JS crypto isn't particular safe either (there are realistic side-channel attacks on it), it's not the cluster•••• that content-controlled JS crypto is.
There are much, much bigger problems with Cryptocat than Javascript. I would also recommend avoiding the mobile apps, which are native.
The scorecard lists "good features" and gives you a tick if you claim to offer that feature. The score card does not currently test any of those and certainly hasn't tested the auditing or done their own audit.