Aug 022015

It’s been a while since my last post – and this one is a doozey.

So Bind is one of the most popular DNS servers on the planet. Just about everyone runs it. So when news breaks that a specially crafted request can cause the named process to exit, then a problem is presented.

Enter CVE-2015-5477.

The official report says:

named in ISC BIND 9.x before 9.9.7-P2 and 9.10.x before 9.10.2-P3 allows remote attackers to cause a denial of service (REQUIRE assertion failure and daemon exit) via TKEY queries.

This doesn’t really convey the severity of the issue. Thankfully, the ISC elaborate more. In it, they say:

The practical effect of this is that this bug is difficult to defend against (except by patching, which is completely effective) and will not be particularly difficult to reverse-engineer. I have already been told by one expert that they have successfully reverse-engineered an attack kit from what has been divulged and from analyzing the code changes, and while I have complete confidence that the individual who told me this is not intending to use his kit in a malicious manner, there are others who will do so who may not be far behind. Please take steps to patch or download a secure version immediately.
This bug is designated “Critical” and it deserves that designation.

Essentially, “You’re screwed. Upgrade now”.

If you’re a system admin, and you’re reading this, check your bind version now, make a coffee, then dig in for the long haul.

Jan 312015

So the openjdk in most linux distros has now been upgraded to v1.8. This has a good bug fix regarding the whole SSLv3 Poodle vulnerability.

This has one problem. The Dell DRAC remote management cards installed in a lot of Dell servers relies on SSLv3 to operate. Without this, you can get into the web interface – but when you get an error stating Error when reading from SSL socket connection and no further.


Thankfully, it is simple to re-enable SSLv3 to allow the connection to succeed.

Open up /usr/lib/jvm/*/jre/lib/security/ in your favourite editor as root, and change the following line:



This enables SSLv3 to all java applications – however it exposes yourself to the MITM attack as defined in CVE-2014-3566. I suggest having a read of the CVE to understand if you want to leave this setting as default on your system or disable it again afterwards.

Jan 292015

A while ago I wrote about how to do this exact thing but with an older version of openssh.

If you’re running a newer version of SSH, then the command syntax has been updated somewhat.

Firstly, once you’ve got your yubikey, you’ll need to enable EPEL for EL6/7 and install the pam_yubico package.

You’ll then need to modify the sshd pam file /etc/pam.d/sshd. There are two options here.
1) You require just the OTP; or
2) You want the OTP and a password.

If you want just the OTP, you add this just after the #%PAM-1.0 header:
auth sufficient id=16 authfile=/etc/yubikey_mappings

If you want both password AND OTP, you add this:
auth required id=16 authfile=/etc/yubikey_mappings

Now to create the /etc/yubikey_mappings user to key mapping. The README says:

Create a /etc/yubikey_mappings, the file must contain a user name and the
Yubikey token ID separated by colons (same format as the passwd file) for
each user you want to allow onto the system using a Yubikey.

The mappings should look like this, one per line:

first user name:yubikey token ID1:yubikey token ID2:….
second user name:yubikey token ID3:yubikey token ID4:….

Now, if you want to go further and require both a ssh key AND an OTP, you can add the following to /etc/ssh/sshd_config:
AuthenticationMethods publickey,password

Now after you supply a valid ssh key you will be asked for your password. If you’ve set this up correctly, this will either be your password + OTP or just OTP.


Update 21/Jun/2015
One common question I get is how they can allow access without a yubikey while in the office, but force its usage outside of the office. This has a couple of parts – mainly, you’ll probably want to use a public key from inside, but force say a publickey + yubikey outside.

We do this by using a Match block in /etc/ssh/sshd_config as follows:
AuthenticationMethods publickey,keyboard-interactive
Match Address
AuthenticationMethods publickey

In this method, we set that EVERYONE must use a public key and a keyboard-interactive method to authenticate, then we allow exceptions for small address spaces that we trust. I also recommend making the following changes:
PasswordAuthentication no
ChallengeResponseAuthentication yes

This disallows skipping the yubikey auth and just using a password. Although, now we’re using PAM as the auth source, you can *still* use a password via PAM – so we need to disable this in /etc/pam.d/sshd:
#auth substack password-auth
#password include password-auth

Hope this helps.

Jan 082015

So I’ve been a bit paranoid of late when reading of the actions of the NSA – and looking at the default configs of sshd that ship with distros like EL6, there is a lot that can be done – however it requires updating to a newer openssh version than the ones that ship with EL6.

I now build openssh (currently v6.7p1) in my testing repo:

After installing this, I use the following to change options as required for ‘best practices’. A lot of these come from here. There is a bit more discussion on this by Aaron Toponce.

Firstly, remove existing SSH server keys and only create the following two. Also set AUTOCREATE_SERVER_KEYS=NO in /etc/sysconfig/sshd to stop missing keys being automatically recreated on start.
cd /etc/ssh/
rm -f ssh_host_*key*
echo AUTOCREATE_SERVER_KEYS=NO > /etc/sysconfig/sshd
ssh-keygen -t ed25519 -f ssh_host_ed25519_key < /dev/null ssh-keygen -t rsa -b 16384 -f ssh_host_rsa_key < /dev/null

Then add some config to /etc/sshd/sshd_config. If you have any Match blocks, this needs to come before them. If not, add the following to /etc/sshd/sshd_config:
## Change key exchange preferences to pick secure methods.
HostKey /etc/ssh/ssh_host_ed25519_key
HostKey /etc/ssh/ssh_host_rsa_key

Then eventually restart the sshd service:
service sshd restart

Remember to always keep an SSH session open to a server as you do these – as if you get it wrong, a failed start of sshd may lock you out of that system!