Heartbleed bug

[_jr_]

Actually ordinary users have to do quite a lot for quite some time. Right now there are half a million possibly/probably (take your bet) compromised standard CA signed certificates in the wild. The only way to stay secure until the neccessary revocation has happened is to remove trust from each and every root CA and just accept the individual certificates after thorough inspection. All the fancy colors the browsers display for secure connections are worth nothing right now. Actually it is more secure right now to accept a newly created self signed certificate deployed on a fixed system (even though the browser actively tries to discourage you from doing this) than a standard CA signed one older than a couple of days.

 

[comradekingu]

Think im unpatched on my phoner come to think of it

 

[DrCJBoduma]

Spoiler

 

[levi]

The people who should do something ASAP isn't ordinary users - It is server admins. The big catch isn't individual users passwords, but the servers key pairs and certificate. With that, any malicious intruder can mount a man-in-the-middle without even having to do SSL-strip or anything like that.


For good info and links: https://www.schneier.com/blog/archives/2014/04/heartbleed.html

All systems with recent enough implementations of OpenSSL that haven't been patched are vulnerable. Granted, your PC might not be running a server processes protected by TLS, but even so it will respond to heartbeat requests sent by a server it connects to. Say the certificate for a big site like google got compromised, and someone managed to perform a man-in-the-middle attack and appear as that big site. Once you connect to it it can fire off dozens of malformed heartbeats to pull SSL/TLS data off your device before anything times out, or if it even pretends to be the real server, much longer, and you'll have entered in your login credentials by then. Theoretically your SSL/TLS data could include symmetric keys used to encrypt recent transactions which could be retroactively decrypted.

As _jr_ says this makes HTTPS X.509 certificates from a CA no more worthwhile than a self-signed X.509 certificate, and actively misleading in the worst case. SSL/TLS still protects your communications against snopping between source and target same as it always has, but if the target can be spoofed by other agencies you're not guaranteed against further attacks.


As far as I can tell, other common uses like SSH aren't affected by this spoofing since they don't use X.509 keys, but are still open to this bug in other ways like pulling off other keys which might be useful to an attacker.


As for my phone, I think it's okay as it's not new enough, and not had a recent enough build of the OS to include any heartbeat code, defective or not.

 

[Magic Sam]

Hi,

OpenSSL has been forked here.

Source code (only for OpenBSD at the moment though)

Dev blog

Cheers,

Magic Sam

 

[WizardStan]

But why? From the website, it claims "LibreSSL is a FREE version of the SSL/TLS protocol forked from OpenSSL" with major emphasis on the "FREE", but OpenSSL was already under a classified free (as in both freedom and beer) license: that's why they were able to fork it at all. This isn't a case of OpenOffice vs LibreOffice where trademarks got borked and a fork and new name was required for true freedom, as near as I can tell OpenSSL is completely free in every sense. Or did I miss something?

 

[levi]

Because Theo had a hissy fit about the developers of OpenSSL and took his ball away to another playground.

 

[WizardStan]

Yeah, reading about how patches have difficulty getting accepted (heartbleed apparently still hasn't been patched in main, so many weeks later) I can understand the desire to locally fork, but to try and make a competitive fork, and then advertise it as the "free" version (giving it the "libre" moniker no less) makes me question things... Something feels wrong about this and I can't really place my finger on it.

 

[levi]

Well, Theo traditionally uses the 'open' prefix - OpenBSB, OpenSSH etc., but unfortunately OpenSSL was already taken as a name.  Plus it seems to have many names, not just LibreSSL - also libssl as the github project name, and this bizarre valhalla rampage stuff on the dev blog.  The media seem to have settled on LibreSSL though, so perhaps that will stick.

As for his decision to fork it, he's got plenty of experience running projects like this and organising code reviews, which is exactly what this code needs at the moment, so I say more power to him.

In other news, Apple have announced that their routers and time capsule backup stuff is vulnerable to heartbleed, since it uses OpenSSL.  Their MacOS and iOS stuff uses their own TLS implementations, so suffers from different bugs.

 

[Magic Sam]

[bump]

Hi,

LibreSSL is ready for porting

Cheers,

Magic Sam

[/bump]