Awesome find by https://hackerone.com/joaxcar and it's a little bit frightening that this severe of a bug isn't the biggest security vulnerability of today
This seems super light on details and I had to dig to find more information. We don't use Gitlab, but we do use Ruby, so I was worried we might be impacted.
So the problem is in Ruby itself? Not in onigmo, or in oniguruma? If simply updating Ruby fixes the problem... is GitLab vendoring a Ruby runtime or something? Or has Ruby not published this fix yet?
… Or is the GitLab/onigmo bug about failing to escape regex input?
Yes. Omnibus ships its own Ruby under /opt/gitlab/embedded, which is why the changelog has "Add Ruby regexp repeat ID overflow patch for 3.3, 3.4 and 4.0" in every recent branch. So patching your distro Ruby does nothing for GitLab, and GitLab's patch does nothing for your other Ruby apps.
To confirm what you're actually running after the upgrade, check /opt/gitlab/embedded/bin/ruby -v on each node. Also look at the two detections GitLab published (Onigmo native worker crash, Org mode render error burst) and run them back over your logs from before you patched.
This is shaping up to be the single worst security news day ever. Given the role that AI played in some of the discoveries, one has to wonder if this marks the start of some dark age of IT security.
I think this marks the beginning of the golden age. Investments in automated agent driven security scans on all software is coming to every business. Software and infrastructure will become more secure than ever before.
I also hope that we come out in a golden age, but we are very much in for a rough few months/years where all the low hanging fruit is being picked up until we end up with much more hardened systems.
Maybe I'm not enough in the weeds to see it as much, but the last few months seem to have been a much slower trickle of "how did we not catch this" exploits, while they are now appearing with much more sophisticated (though still usable in practice) exploit chains on much larger target codebases.
The last few months seem to have been a lot more imunder control with also the frontier labs more warning of "what happens when Mythos-class model get in the hands of everyone" than that actually becoming reality. Now, it appears to be actuually turning into reality.
The problem is fixing the bug and not only fixing the code, but applying the fix in production in a timely manner and without interrupting the business.
It's ok, we'll just point more LLMs at the new code and the problem will fix itself. In the mean time can we unplug the ethernet cable from the nuke silos for a while?
Code does not need to be reviewed to be secure. [^]
It needs to be designed to be secure.
[^] of course I'm hyperbolic here, it does still need to be reviewed – I just mean, review doesn't matter as much as choosing a secure design, where certain bugs just can not or will not happen.
Code review is just one of many filters. 'Secure designs' are as secure as the designer, and no designer is perfect. Implementations are as secure as the implementors and processes are not perfect either. You need these and many other layers in your filter to be secure for some arbitrary level of security where the final result is roughly proportional to how much money you are willing to expend.
At least second regex related CVE in GitLab in the past few weeks. Awful to see that the endpoint is reachable unauthenticated, but I'm glad that they provide guidance in how to check the logs for attempts of exploitation
As far as I can tell, it requires forcing GitLab to render a malicious org-mode file. I don't really see a way to do that without authenticating and uploading one. Am I missing something?
Back in perl 5.8 and older, the regex engine worked by recusing at the level of C functions, and so you could trigger stack overflows by giving it large strings.
Is this similar?
I wonder why gitlab even has its own regex compiler. Or is this a vulnerability in an underlying library or programming language?
The downvoters must have missed Airplane!! For a second I thought about how to escape the ! to make it clear it’s part of the title… but I haven’t had any coffee.
Gitlab is probably more prominent since it's a large enterprise piece of software that's self hosted. Every bug in a library used by Gitlab becomes a problem for Gitlab. While hosted platforms like Github can just update the library and not have to make an announcement.
Years ago I worked at a company that had a nightly cronjob for yum to apply security updates (with a couple of important exceptions that were handled manually). Always seemed very iffy to me for such a big company where downtime was very expensive.
No. Like the PHP vulnerability, this was a bug in a language interpreter written in C, and a bug that would not be possible in most newer languages. Even as secure-by-design becomes more prevalent, there will be vulnerabilities, but this one is a taste of the past.
GitLab has remediated an issue that, under certain conditions, could have allowed an unauthenticated user to execute arbitrary code on the GitLab server due to integer overflow when compiling a specially crafted regular expression. (...)
Truth be told, I don't think that "GitLab Critical Patch Release: 19.5.1, 19.4.3, 19.3.6, 19.2.9, 19.1.9, 19.0.10, 18.11.13" is a more descriptive title, even taking the HN guidelines into account. It could be better, sure, but it's not horrible (IMO).
It's an editorialization. The submitter is saying something about the importance of CVSS that the article does not. A better title would say "Remote Code Execution" where it says "CVSS 10.0" now.
I still remember a supposed 9.9 bug in CUPS (CVE-2024-47176) that was effectively irrelevant for almost everyone. Basically, as I understand it, you'd have to be running a server that let people connect from the public Internet that was also running CUPS for some reason, and then also have someone inside the organization try to print something to a fake attacker-controlled "printer". But it would have been RCE, yes.
My experience has been that CVSS has really just not been very informative or well-calibrated to risk.
No? CVSS is a Ouija board. Citrix directory traversal got scored 10.0. So did a Cisco Broadworks SSO bypass. Meanwhile, Shellshock did not.
CVSS is like the Pitchfork Music ratings scale. It communicates false precision, and it's easy to point to places where that precision misleads (D-plan's "Emergency & I" is not better than every Yo La Tengo album).
Have you read signal/noise theory? That a signal isn't perfect (that it contains false positives or positive falses in case of binary signals, or that it has a delta between the signal and the underlying phenomenon) doesn't mean that it carries no information.
For example if I cry wolf 50% of the time, but 100% of the times there's a wolf I cry wolf. Then no matter what the frequency of wolf sightings are, my crying wolf carries some information (unless there's always a wolf). Assume there's a wolf 1% of the nights, then I will cry wolf 50% of the time, and you only need to check for a wolf 50% of the nights rather than 100%.
It's worth noting that there's still technically signal even if I sometimes miss a wolf, even if you may choose to ignore that signal in order to pursue 100% accuracy yourself.
It's more like: you work as an animal spotter, and if you tell people you saw a hamster, you'll get no glory. So obviously you spotted a direwolf at least 12 feet tall.
Interestingly CVSS was designed to help with this specific issue. You are complaining that the system is imperfect, but you fail to imagine how it works without such a system, everyone tagging their issue as critical.
I am concerned for my self hosted servers, well only hope is tailscale should be entirely secure, and egress is filtered heavily... If that doesn't work my home lab is fucked.
Any suggestions folks, not even my terminally online self can handle all of these.
This is the best overview I found: https://lilting.ch/en/articles/gitlab-cve-2026-110264-org-mo...
The bottom line: this is an issue in Ruby's regex engine, which was fixed in April and backported recently. [1]
Everyone running applications where attacker-controlled input might find its way (unescaped) into Ruby regexes is vulnerable to remote code execution.
[1]: https://github.com/ruby/ruby/pull/16801
… Or is the GitLab/onigmo bug about failing to escape regex input?
Big malicious entities always had more investments in this than random open source projects.
The last few months seem to have been a lot more imunder control with also the frontier labs more warning of "what happens when Mythos-class model get in the hands of everyone" than that actually becoming reality. Now, it appears to be actuually turning into reality.
I'd say, the single worst security news day for now
It's over. Cybersecurity is now Yugioh, where the lifepoints is your token budget.
Can I get that in writing?
As we speak more slop is added to the pile, far more than can reasonably be reviewed and a lot of that code will be insecure.
It needs to be designed to be secure.
[^] of course I'm hyperbolic here, it does still need to be reviewed – I just mean, review doesn't matter as much as choosing a secure design, where certain bugs just can not or will not happen.
Is this similar?
I wonder why gitlab even has its own regex compiler. Or is this a vulnerability in an underlying library or programming language?
after updating a pop-up still shows:
"You are currently on version 19.5.1! We strongly recommend upgrading your GitLab installation immediately."
while 19.5.1 should be the latest and fixed version
But I also updated pretty early.
Coincidentally, there is also a (non-embargoed) musl libc CVE published today involving integer overflow in its regex compiler: https://www.openwall.com/lists/musl/2026/10/21/1
Perhaps that is the underlying bug, or a similar one turned up by the same security research?
GitLab has remediated an issue that, under certain conditions, could have allowed an unauthenticated user to execute arbitrary code on the GitLab server due to integer overflow when compiling a specially crafted regular expression. (...)
CVSS 10.0 (CVSS:3.1/AV:N/AC:L/PR:N/UI:N/S:C/C:H/I:H/A:H)
Here, CVSS 10.0 means arbitrary remote code execution.
My experience has been that CVSS has really just not been very informative or well-calibrated to risk.
Might you perchance be exaggerating?
CVSS is like the Pitchfork Music ratings scale. It communicates false precision, and it's easy to point to places where that precision misleads (D-plan's "Emergency & I" is not better than every Yo La Tengo album).
For example if I cry wolf 50% of the time, but 100% of the times there's a wolf I cry wolf. Then no matter what the frequency of wolf sightings are, my crying wolf carries some information (unless there's always a wolf). Assume there's a wolf 1% of the nights, then I will cry wolf 50% of the time, and you only need to check for a wolf 50% of the nights rather than 100%.
It's worth noting that there's still technically signal even if I sometimes miss a wolf, even if you may choose to ignore that signal in order to pursue 100% accuracy yourself.
CVSS is good, it's better than nothing.
Any suggestions folks, not even my terminally online self can handle all of these.
But it's a lot better than having things live on the open internet!
Also looking at those old C++ services a whole lot differently now.