SSH client library exploitable by MITM or by malicious server.
Versions to 1.11.1 likely vulnerable. Keep an eye out for incoming update!
https://www.cve.org/CVERecord?id=CVE-2026-55200 or NVD - cve-2026-55200
I’m unsure of whether it is being exploited in the wild but a PoC is public so it can only be a matter of time.
what are the poc requirements for mitm? I think I was a lazy man and didnt upgrade one of my byzantiums to crimson yet altho I may do so this weekend.
I routinely ssh across the ocean to VPS and such, is this putting me at risk to any router along the way for MITM ?
I haven’t inspected the PoC but the main point is that if a client is vulnerable to a maliciously crafted packet before authentication occurs (in this case, authentication of the server to the client) then this attack will work as MITM.
To be honest though, the CVE itself doesn’t say that explicitly but I have seen it written elsewhere. Unless you are really really familiar with the details of the SSH2 protocol, I would not assume that this can’t be done MITM.
You don’t have to worry about byzantium v. crimson right now because, as far as I know, at the time of my post yesterday, there is no patch released for any Debian family operating system.
Thanks this is good information.
Do you know if SSH servers are vulnerable too? I have usually assumed that the server is a security hole on Librem 5 because I might get on an airport wifi and get hammered by someone trying to SSH into me. Also if I ever am installing a tool or something and install the sshd server as a dependency or for a project, the password is the insecure pin and the user is always purism so it’s very much asking to be hacked.
But my solution was usually to proactively install the sshd server to the Librem 5 in case I ever might and then DISABLE the password authentication, without adding any ssh keys so nothing can ever actually log in.
This means there is a live sshd server with me often in my pocket. Can I walk past someone who does RCE now with this by abusing the same bug in the server code? Or is it only client?
Edit: Perhaps I misremembered… Looks like this is disabled?
purism@pureos:~$ sudo systemctl status ssh.service
● ssh.service - OpenBSD Secure Shell server
Loaded: loaded (/lib/systemd/system/ssh.service; disabled; vendor preset: enabled)
Active: inactive (dead)
Docs: man:sshd(8)
man:sshd_config(5)
purism@pureos:~$ ssh localhost
ssh: connect to host localhost port 22: Connection refused
purism@pureos:~$
purism@pureos:~$ sudo systemctl status ssh.service
● ssh.service - OpenBSD Secure Shell server
Loaded: loaded (/lib/systemd/system/ssh.service; disabled; vendor preset: enabled)
Active: inactive (dead)
Docs: man:sshd(8)
man:sshd_config(5)
libssh2 is only used in clients, not servers.
As long as you only connect to hosts already in your known hosts, you should be safe as the host key would change in any MitM scenario.
In case an attacker was able to compromise an SSH server you connect to, this could result in a compromised client without any notice I think.
For public WiFis you should never expose any services on your device. They may have client isolation or not and you never know how trustworthy the network is. You may think about an always-on Wireguard VPN to your home network.
Wouldnt this be false because the communication layer discussing the host key would itself be the exploitable program?
Ah thanks for pointing out. I looked a bit further and in fact the OOB triggers before host key check.
But I would nevertheless point out it’s not “very easy” to get RCE in a real life scenario. Not that I want to play it down but there would be further prerequisites, e.g. on modern systems which mostly implement ASLR.
To my mind, Red Hat has a nice short overview of the vuln: cve-details
Sorry, I don’t know.
However, as a consequence of this vulnerability, you would expect that some people will look at similar code on the server side, even if the only people doing that are the black-hats. 
Let’s suppose, as a hypothetical, that an SSH server were vulnerable to a similar exploit …
… that really isn’t up to the job - based on the discussion in this topic about pre-authentication exploits. Right?
You always have to make your choices but one choice would be: install and configure the SSH server but then don’t start it. You will usually be able to start the SSH server manually from the phone if you need the SSH server.
The downside of that choice is that maybe something is screwy with the UI and so you can’t start the SSH server manually but you have the expectation that the problem that you are experiencing was such that an automatically started SSH server would have started successfully, and would be useful in diagnosing that problem.
Per CVE-2026-55200 both byzantium and crimson could be vulnerable.
No its about clients. Like your Server (like in Wordpress) got nudge to interact with a unknown evil new Ip (or customer) to interact with ssh. It could be attacked. I think your Phone is not likely to be a target.
For Servers try to compile libssh2 with tthe Patch and you are fine. If you do not need some public on your systems just try to nail down Firewalls rules to only talk to your systems like port knocking with a daily biases and you are fine.