Nov 142001

ssh exploit – how to avoid it

Over IRC, I’ve learned about an SSH exploit. And I’ve also learned how to avoid the exploit. I’m not claiming to be an expert in this area. I do claim to be relaying information which might help you. Those amongst you that know more about this area than I will hopefully add their comments to this article.

What’s the exploit?

This vulnerability is described in CERT Vulnerability note VU#945216. Their description of the problem is:
There is a remote integer overflow vulnerability in several implementations of the SSH1 protocol that allows an attacker to execute arbitrary code with the privileges of the SSH daemon, typically root.
The details of the analysis are at

The FreeBSD Security advisory is at

So what isn’t vulnerable?

My understanding is that FreeBSD 4.3 onwards is not vulnerable to the SSH1 exploit. Even if you have an older FreeBSD, you can avoid the exploit by using SSH2 instead of SSH1. How do you do that? Read on…

Of course you could merely upgrade your FreeBSD to the latest version. Or you could install the OpenSSH port.

Using ssh2

Given that SSH1 is vulnerable, how can you reduce your exposure? First, you need to realise that it is still very difficult to exploit the SSH1 bug. Second, some clients are not SSH2 capable. That said, I think there are a few simple steps you can undertake in order to reduce your exposure. The easiest step is to ensure that both your clients and your servers use SSH2 instead of SSH1, if it’s available. How to do that is outlined below. A more dramatic step would be to require SSH2. But that’s not for everyone.

Change your servers

You can specify the protocols used by your sshd via /etc/ssh/sshd_config. Look for the entry which looks like this:
Protocol 1,2

That specifies that both protocols 1 and 2 should be accepted. It does not determine which protocol will be tried first. That decision resides with the client. The following is from man sshd:

Specifies the protocol versions sshd should support. The possible values are `1'' and `2''. Multiple versions must be comma-separated. The default is `1''.

If you wish to use only SSH2, that is an option. But consider it carefully. The SSH1 exploit is difficult to achieve. Some clients are not SSH2 compatible. If you can ensure that all your clients will be SSH2-compatible, then, and only then, is SSH2 only an option for you. If so, adjust your configuration file to be this:

Protocol 2

WARNING: Use the above with caution considering who your clients will be.

Change your clients

The change to your clients are quite similar to your sever changes. The file to modify is ~/.ssh/config (which might not exist on your system; you may need to create it). I added the following line:
Protocol 2,1

The following is from man ssh:

Specifies the protocol versions ssh should support in order of preference. The possible values are ''1'' and ''2''. Multiple versions must be comma-separated. The default is ''1,2''. This means that ssh tries version 1 and falls back to version 2 if version 1 is not available.
Note that order is significant. In the example above, Protocol 2 will be tried and then protocol 1. You can also specify protocol 2 using the command line:
ssh -2


ssh -o "Protocol 2,1"

Create and use SSH2/DSA keys

The following two resources are compulsory reading if you are using SSH. It describes how to properly manage OpenSSH keys. It will also show you how to use the more advanced features of OpenSSH.

You can create SSH/DSA keys using ssh-keygen -d. Definitely specify a passphrase. Reading the above URLs will show you how to use ssh-agent to reduce the number of times you need to enter the passphrase. It’s very useful. Your public DSA key (~/.ssh/ needs to be added to ~/.ssh/authorized_keys2 on the remote box (i.e. the sshd server). Read the above URLs for complete information.

  2 Responses to “ssh exploit – how to avoid it”

  1. Whilst I think of it, for those of us that happen to use X in everyday environment here is an idea for everyone, if you are consistentley sshing to machines and get sick of putting in your password each time in your .xsession or .xinitrc put in the following…

    eval `ssh-agent -s`
    /usr/bin/ssh-add ~/.ssh/identity < /dev/null
    /usr/bin/ssh-add ~/.ssh/id_dsa < /dev/null

    (For both ssh 1 & 2)

    When X starts up it will prompt you for passwords for both of these, put them in and ssh to a machine that already has your keys in the authorized_keys or authorized_keys2 and provided everything went well you will be logged in with no need for a password. It really is handy just ensure to lock your workstation when you walk away for obvious security reasons.



  2. Hi,

    although someone seems to propagate that version 1 is insecure in
    general it’s only an implementation bug. Therer are patches available for various versions of OpenSSH for *BSD, Linux and some oder platforms. Look at for details.

    Greetings, Marcel