Please clarify – will the update add *Roam* to /etc/ssh/ssh_config? I’ve just checked on two systems that are CentOS 7, a server, and a workstation that I literally built yesterday, and grep -i on both reports “no, not here”.
mark
It will fix the bug.
Yes, as it’s undocumented, but enabled since about 2010. Even OpenBSD
5.9 (pre-release, it’s going to be released on May 1st, 2016) does not mention it.
That came from Theo (OpenBSD’s Theo) and was called undocumented. So, my guess is that, in the client (not the server) there is a default of UseRoaming that doesn’t show in the config file.
Note that this is something that affects SSH clients, not servers.
Timo Schöler wrote:
Undocumented? You’re saying that there’s a feature that is configurable via the configuration file, and there’s no mention of it at all in the configuration file, not even the default?
That is more than slightly unacceptable.
mark
More than agree! I was highly respecting OpenBSD project, especially for their openssh. After scandal with OpenBSD IPSEC stack backdoor accusations, my respect faded grossly, and I felt extremely happy my choice of system for servers fell on FreeBSD, not OpenBSD (for some independent reason)…
Valeri
++++++++++++++++++++++++++++++++++++++++
Valeri Galtsev Sr System Administrator Department of Astronomy and Astrophysics Kavli Institute for Cosmological Physics University of Chicago Phone: 773-702-4247
++++++++++++++++++++++++++++++++++++++++
RH issued an update to address this a bit over an hour ago:
Yes, thank you, I saw it yesterday in my e-mail from yum.
I am not happy that this bug existed, undocumented features enabled by default are not a good thing.
However that this bug was found demonstrates a success of the Open Source philosophy. I don’t know this would have been found in a closed source SSH implementation.
Open Source works.
I see that this is a CentOS 7 patch only, at least so far. I also see that the CentOS 6 SSH version is 5.3
> /usr/bin/ssh -V
OpenSSH_5.3p1, OpenSSL 1.0.1e-fips 11 Feb 2013
which is supposedly not affected. However, strings indicates that /usr/bin/ssh is also aware for the useroaming configuration option:
> strings /usr/bin/ssh | grep -i useroam useroaming Is it actually known that the SSH version shipped with CentOS 6 is not vulnerable, or is it just assumed based on the version number? The announcement implies that the roaming code itself was added in 5.4, not just that a default was changed, but if that
Alice Wonder wrote:
Complete agreement. Yup. Certain closed-source coMpanie$ would be saying “what problem, there’s no problem here, pay your money and move along….”
First, because versions 5.4 – 5.6 were not vulnerable to the information leak on GNU/Linux, though they were on BSD systems. Second, because later versions may have been able to leak private keys, but only incomplete copies of them. Last, because encrypted keys could only be leaked in their encrypted form, and keys used with an ssh-agent were not vulnerable to leaking at all.
The buffer overflow vulnerability seems more severe, but only if you’re using a bastion host which is compromised. The vulnerability can only be triggered when using ProxyCommand. The buffer overflow also is not exploitable on OpenSSH 6.8, due to a bug introduced in that version.
15 thoughts on - Heads Up: OpenSSH Users
Michael H wrote:
Please clarify – will the update add *Roam* to /etc/ssh/ssh_config? I’ve just checked on two systems that are CentOS 7, a server, and a workstation that I literally built yesterday, and grep -i on both reports “no, not here”.
mark
It will fix the bug.
Yes, as it’s undocumented, but enabled since about 2010. Even OpenBSD
5.9 (pre-release, it’s going to be released on May 1st, 2016) does not mention it.
Timo
—–BEGIN PGP SIGNATURE—
FYI:
https://github.com/openssh/openssh-portable/search?q=AppGate+Network+Sec urity+AB
—–BEGIN PGP SIGNATURE—
That came from Theo (OpenBSD’s Theo) and was called undocumented. So, my guess is that, in the client (not the server) there is a default of UseRoaming that doesn’t show in the config file.
Note that this is something that affects SSH clients, not servers.
Timo Schöler wrote:
Undocumented? You’re saying that there’s a feature that is configurable via the configuration file, and there’s no mention of it at all in the configuration file, not even the default?
That is more than slightly unacceptable.
mark
More than agree! I was highly respecting OpenBSD project, especially for their openssh. After scandal with OpenBSD IPSEC stack backdoor accusations, my respect faded grossly, and I felt extremely happy my choice of system for servers fell on FreeBSD, not OpenBSD (for some independent reason)…
Valeri
++++++++++++++++++++++++++++++++++++++++
Valeri Galtsev Sr System Administrator Department of Astronomy and Astrophysics Kavli Institute for Cosmological Physics University of Chicago Phone: 773-702-4247
++++++++++++++++++++++++++++++++++++++++
RH issued an update to address this a bit over an hour ago:
<https://rhn.redhat.com/errata/RHSA-2016-0043.html>
I expect that we’ll see the CentOS version shortly.
In what way do spurious accusations reflect on the accused, instead of the accuser?
I do agree that leaving code for an undocumented feature in the client was irresponsible, but that’s an entirely different matter.
For the record, this update is now released (it was yesterday):
https://lists.CentOS.org/pipermail/CentOS-announce/2016-January/021614.html
This contains a patch that disables roaming:
https://git.CentOS.org/commitdiff/rpms!openssh.git/1edce7e6bfedb27a163f35bcacab620a703408ac
Thanks, Johnny Hughes
Yes, thank you, I saw it yesterday in my e-mail from yum.
I am not happy that this bug existed, undocumented features enabled by default are not a good thing.
However that this bug was found demonstrates a success of the Open Source philosophy. I don’t know this would have been found in a closed source SSH implementation.
Open Source works.
I see that this is a CentOS 7 patch only, at least so far. I also see that the CentOS 6 SSH version is 5.3
> /usr/bin/ssh -V
OpenSSH_5.3p1, OpenSSL 1.0.1e-fips 11 Feb 2013
which is supposedly not affected. However, strings indicates that /usr/bin/ssh is also aware for the useroaming configuration option:
> strings /usr/bin/ssh | grep -i useroam useroaming Is it actually known that the SSH version shipped with CentOS 6 is not vulnerable, or is it just assumed based on the version number? The announcement implies that the roaming code itself was added in 5.4, not just that a default was changed, but if that
Alice Wonder wrote:
Complete agreement. Yup. Certain closed-source coMpanie$ would be saying “what problem, there’s no problem here, pay your money and move along….”
mark
https://bugzilla.redhat.com/show_bug.cgi?id=1298032#c16
(see comment 16)
Yes, that answers my question. Thanks.
Noam
For the sake of conversation…
Reading the Qualys security advisory is interesting as well, and I
tend to think the vulnerability is not severe for a number of reasons:
https://www.qualys.com/2016/01/14/cve-2016-0777-cve-2016-0778/openssh-cve-2016-0777-cve-2016-0778.txt
First, because versions 5.4 – 5.6 were not vulnerable to the information leak on GNU/Linux, though they were on BSD systems. Second, because later versions may have been able to leak private keys, but only incomplete copies of them. Last, because encrypted keys could only be leaked in their encrypted form, and keys used with an ssh-agent were not vulnerable to leaking at all.
The buffer overflow vulnerability seems more severe, but only if you’re using a bastion host which is compromised. The vulnerability can only be triggered when using ProxyCommand. The buffer overflow also is not exploitable on OpenSSH 6.8, due to a bug introduced in that version.