An analysis of the Keycloak authentication system

What if instead of publicly blaming an OSS product, you try to get a support contract with some of the engineers behind it? If your company is too cheap for that, maybe a PR would have been nice?

Having very high expectations when using the software without contributing anything else than public shaming on something that clearly state in the license: "Licensor provides the Work ... WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND" shouldn't be ok, this is quite literally how you make open source developer to burn out

> What if instead of publicly blaming an OSS product, you try to get a support contract with some of the engineers behind it? If your company is too cheap for that, maybe a PR would have been nice?

Yeah, no. That's not how security research works.

If I disclose a security issue to you, it doesn't matter if you're a multinational trillion dollar corporation or a hobbyist in Nebraska, the onus is on you to fix it. Not the security researcher. Their job is done once it's disclosed.

From the timeline:

> 28/03/2024 – First communication sent with all details and a proposed fix.

After that point, any additional help (including a pull request) is going above and beyond.

I run into this attitude you're exhibiting a lot. Where proprietary software has the legal threats, the open source community is plagued by patch entitlement.

Knowledge of a security issue in a project is, in and of itself, a valuable contribution. Expecting a PR devalues this work.

Money doesn't guarantee anything, as seen by the recent issue with Okta. If you're using a centralized auth service and expecting no vulns or exploits, you're going to be severely disappointed. These are the juiciest targets and very complex pieces of software.

I don't see an issue with calling out this Keycloak response time either. It's not inherently negative, just stating facts.

> If I disclose a security issue to you, it doesn't matter if you're a multinational trillion dollar corporation or a hobbyist in Nebraska, the onus is on you to fix it. Not the security researcher. Their job is done once it's disclosed.

On the other hand, if I'm a hobbyist, I have 0 obligations to do or fix anything I've made open source. Patches are welcome ofcourse.

Keycloak is a Red Hat product and is a dependency for many Red Hat products so I'd love it if people running the open source release can report the bug and get feedback. This isn't a student eating ramen supporting this software, its IBM.

>I have 0 obligations to do or fix anything I've made open source.

While technically true, this seems pretty scummy when you're advertising security software for real people and companies to use as their identity management.

Nowhere on the Keycloak home page does it say "just a hobby project" or anything that would remotely indicate that it is not a serious project and that you shouldn't use the software.

Instead, it seems like they are trying very hard to be taken seriously as an identity management product.

There's no obligation for a hobbyist to fix the stuff they publish online. If they're selling the result, then sure, you can argue there's reasonable consideration, but just because a security researcher has made a contribution (and a valuable one) it doesn't compel any further contribution from the original author. Now, the lack of action does probably remove some of the credibility of the project as one that should be used in any security context (bit of a problem for something intended to be used as authentication).

> Nowhere on the Keycloak home page does it say "just a hobby project" or anything that would remotely indicate that it is not a serious project and that you shouldn't use the software.

https://github.com/keycloak/keycloak/blob/main/LICENSE.txt#L...

Indeed

   7. Disclaimer of Warranty. Unless required by applicable law or
      agreed to in writing, Licensor provides the Work (and each
      Contributor provides its Contributions) on an "AS IS" BASIS,
      WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or
      implied, including, without limitation, any warranties or conditions
      of TITLE, NON-INFRINGEMENT, MERCHANTABILITY, or FITNESS FOR A
      PARTICULAR PURPOSE. You are solely responsible for determining the
      appropriateness of using or redistributing the Work and assume any
      risks associated with Your exercise of permissions under this License.

Point 7 buried in the license document of the github repository is very much not https://www.keycloak.org/

> There's no obligation for a hobbyist to fix the stuff they publish online.

Correct, but I didn't say there was. The onus being on them to fix it is predicated on anyone having such an obligation at all. If anyone has an obligation, it's always the vendor, not the researcher.

But keep in mind this situation involves an IBM-funded open source project under the Red Hat product line. The hobbyist remark is tangential to the story.

Keycloak is not written by some random person in the middle of nowhere who is shouldering the burdens of the world. It is security sensitive software that in part is funded by major open-source corporations like Red Hat, including full time engineers. What you're reading in this blog is literally standard fare in security research where a disclosure happens after some set period of time, regardless of when, or even if, the fix happened.

Fixing authentication bypasses is not "very high expectations", it's exactly what you would expect from this software. Get a grip on reality, please.

Customers have been refunded in full.

You really feel that anything that is not directly on the home page does not matter? A link to a separate document explicitly named as containing the conditions of license, warranty and such should not count?

Seems like an absurd view to me.

For all that I think RedHat is not a poor hobbyist and morally at fault. It's just a different matter altogether. The terms under which the software is provided are clearly spelled and in public view. You're just inventing a reason to disregard them.

>You really feel that anything that is not directly on the home page does not matter?

No, that's not what I said.

When it comes to software like this (one of the most important components of your security architecture), hiding the fact that you wont fix vulnerabilities unless you feel like it halfway down in your legalese-filled license is unethical.

I think it's absurd that people are defending this position. We're not talking about a weather app made by Joe Somebody for a weekend project.

Regardless of whether it's a project run by a big company or a guy in a shed, a security-critical project not fixing critical vulns for 10 months is an important bit of information when judging whether you should use it.

It doesn't really matter whether it's someone's fault, or wrong, or whatever - what matters for the user is that using the project is likely unsafe. Sharing that information is a public service.

Keycloak has been donated to CNCF in 2023. So it's not a RH / IBM product anymore.

I would even go as far as say that it never was; Red Hat had their own product called "Red Hat Single Sign On" that was, for some time, based on opensource Keycloak project, but the opensource Keycloak project has existed before RH SSO. And exists now that RH SSO product has been deprecated (retired? Idk what happened).

Red Hat does offer a "Red Hat build of Keycloak" now, and of course Keycloak would not exists in it's current form without Red Hat.

But saying that "Keycloak is a Red Hat product and therefore Red Hat and / or IBM should support it" would be, in my opinion, harmful for the whole opensource movement. If, by being engaged with opensource project, a company risks it's reputation then such company could decide against any engagement, or would engage only if it could keep control of the project / community around it.

Is this the first time you heard of open source licenses or something? This is standard boilerplate, and it's hilarious to think you get to ask for more from a project you're not even contributing to.

As long as you disclose that right-front on your value statement, yeah, you don't have any other obligation.

> maybe a PR would have been nice?

They did.

> 28/03/2024 – First communication sent with all details and a proposed fix.

And for the second one, a fix existed 2 days after first communication, doing a PR doesn't magically make that go faster.

For the third, a fix was done within a month, but once again was delayed, and didn't land until several months later.

- - -

RH does not provide support contracts for any affordable price. CNCF does not provide support contracts at all.

Is there an open source license that doesn't?

I'm not asking for more. I'm saying I think it is scummy to do the bare minimum when you're advertising yourself as a critical piece of security software and encouraging the use of the software in real security-critical applications.

If there's a Red Hat build of Keycloak, and Red Hat products depend on Keycloak, then this vulnerability is present in all of those Red Hat products.

It is right in the license.

So what gives you a right to download and use the software in the first place? The copyright law forbids that by default. What permission other than the license do you have?

If you central auth - you have vulnerabilities If you decentralize auth - you have vulnerabilities. It can lead to shadow credentials as well as credential and auth sprawl.

It is about de-risking your approach. Either approaches work until they don't.

RH SSO was the LTS build of keycloak with business support.

Keycloak doesn't publish hot fixes for previous major versions, and these major versions come out on a very tight release schedule / every few months. So if you didn't want to upgrade all the time, you'd have been forced to use rhsso. And now the red hat keycloak build.

https://github.com/keycloak/keycloak/discussions/25688

That's not what these licenses have come to mean. They're a way to reduce the risk that you'll get sued,

but not any "I don't give a fuck" statement.

You could add "I don't care about fixing security vulnerabilities" somewhere in the beginning of the readme, if you're developing security related OSS software? That'd be more clear.

Maybe the WTFPL actually a little bit indicates that the developers maybe don't give a fuck, though: https://en.wikipedia.org/wiki/WTFPL ?

That sounds a little like having your cake and eating it too. 'Giving a fuck' is not really a boolean value but more of a broad spectrum.

Of course, anyone who writes any software cared a little bit about it at one point, or they wouldn't have written it. But warranty is about whether they care enough to cater specifically to you when you have a problem in the future.

Maybe many of these projects do care enough to give general updates to the community as a whole on a best effort basis, but that's a lower level of assurance and more voluntary than what you'd get in a legal warranty.

> So if you didn't want to upgrade all the time, you'd have been forced to use rhsso.

Or just not upgrade at all. Not the most wise strategy for security-focused software, but I'm sure many teams do that. Especially because keycloak often being heavily customized with plugins and themes, so upgrading this setup might actually be not trivial.

> There's no obligation for a hobbyist to fix the stuff they publish online.

I may be wrong but this might not the case anymore in 2025 (not sure about the timeline) in the European Union because of the new cybersecurity acts.

Off-topic but I love this naming convention from Red Hat which I hope gets more traction across the industry. It absolutely detest wading through vendor marketing material to figure out which open source product is being used under the hood. With names like "Red Hat Build of Keycloak" and "Microsoft Build of OpenJDK" it's crystal clear.

I believe it works out better for the vendors as well because there are so many obstacles with evaluating anything that requires a license in an enterprise setting. If the technical person downloads and evaluates the underlying open source version some manager will insist on purchasing a support contract before going to production.

The number of software engineers who either don’t understand social contracts or only understand them when it benefits them to do so is frankly appalling.

You said you’re going to do something and now people depend on you having done it. Volunteer organizations fall apart if they can’t trust each other.

That was a very articulate and nice way of saying, “what if instead of using the tired old deflection you looked at the facts?”

>You could add "I don't care about fixing security vulnerabilities" somewhere in the beginning of the readme

I care about fixing security vulnerabilities in my OS projects, but I care more about my sanity, my family, getting enough money to survive, and a few other things. Unless you pay me I don't care about your problems with my free (as in a beer) software.

And that's a good thing btw - I tried to ask for donations once, got the equivalent of a few cups of coffee per month, and... burned out almost immediately. I started to feel responsible for that project, staying up late to fix reported minor bugs, and it turns out watching Github issues 365 days a year for a few dollars monthly is not a great business strategy.

Then you should never work on software with security implications, or if you do you should keep it to yourself. I’m a terrible party host, so I don’t host parties. I help other people do so when I can.

I'm not sure what point you are trying to make but thankfully, given that they obviously can't be trusted to maintain security-critical software (despite what they imply on their website and marketing material), I haven't downloaded or used it or recommended it anywhere while consulting.

This is not a one-person project ran by someone in their spare time, posted online for fun.

They are going out of their way to advertise so that people use their security-critical software in security-critical applications, and then they neglect the security.

While they aren't under any legal obligation, it's (in my worldview at least) pretty damn unethical.

All they would have to do to not be unethical is make it clear that this software should not be used in any security-critical application because it is not properly/frequently maintained. Put that in a header on the website.

>a guy in a shed

That should be enough without needing to shame them over timeline.

You can also judge based off issue open/close rate and contribution frequency (GitHub has pretty charts for these if the project is hosted there)

People depending on you is not somethng in your control, therefore you can't be responsible for that. It's up to the people depending on you to make a judgement on whether or not that's a good idea.

Social contracts are not contracts, they do not end up in court, and if you depend on them you have no recourse. The most anyone can say about social contracts is that you are allowed to be disappointed privately. And even then...

I don't have a business presence in the EU, so I rarely care about that for my own projects. (Insofar as I do care, it's limited to "let's not make GDPR compliance logically impossible when designing cryptographic features").

The license is not your value statement.

No, they end up in the court of public opinion. <gestures around>

And they make you lonely if you keep losing.

They are literally explicitly stating that the software does not claim to be fit for purpose.

You can’t have it both ways.

They are advertising on their website, extremely prominently, that they are fit for your all of identity management needs.

Are they allowed to put a single paragraph in their license file that runs counter to all of their other marketing, advertising, and communication efforts? For sure!

Is it shitty to do that? I think so. Just be upfront, it's not hard. If your software isn't fit for security-critical applications, don't pretend it is.

Red hat consumers have been refunded? Where?

As per the terms here: https://www.keycloak.org/pricing

Is it still funded by Red Hat today?

> Point 7 buried in the license document of the github repository is very much not https://www.keycloak.org/

Not a single word in this comment refers to "software like this (one of the most important components of your security architecture)". I guess you realized the absurdity of your position and moved the goalposts to something else.

What?

Can you explain where you think I set goalposts, and how you think I moved them? Because I am not following.

There's carve-outs in that for open source hobbyists ("not associated with commercial activity"). This was originally vaguely worded but they've now made it a lot less ambiguous, the only open source it covers is that which is being developed by a company which is also making money directly from it.

(And in the case that a company takes that code and uses it in a product, they are responsible to fix any security vulnerabilities but also to report it to the author)

The point is that red hat also sells keycloak and develops it. I agree that most users don't pay, but your point is a bit weird considering that some people do actually pay/paid for its development and still do not get a refund

This blog post is from another domain, writing, but it’s stuck with me over the years and colored how I see things like open source projects:

https://journal.neilgaiman.com/2009/05/entitlement-issues.ht...

I suppose it's best if you never use any open source software ever again because they all do that. Like I said, it's standard boilerplate and it's absurd to think you get to wish for more from an open source project that you aren't contributing to.

Of course this boilerplate is necessary else you get people like yourself demanding unreasonable things.

I'm not demanding anything, please stop reading my comments in the most uncharitable way you possibly can.

I'm not sure what has you in super-defense mode, but just as they are allowed to misrepresent themselves on their website, I'm allowed to think that it's shitty to do so. However, as I've said already (a few times, actually), they are more than free to continue doing so (and I'm more than free to keep saying it's shitty).

>[...] because they all do that.

No, they don't all do that.

Several developers and full-time employees, including the project lead, are still employed by Red Hat.

The point is they're not misrepresenting anything, you just don't seem to understand open source. Literally all of them have the same disclaimer, and obviously they're not going to make any guarantees to randos who haven't even paid them. If this is a problem for you, stop using open source. For a start, say byebye to linux.

As for why I'm "on the defensive", bashing open source projects is bad form. You're absolutely welcome to request a refund, though.

The Linux kernel, probably one of the most critically important pieces of software nowadays, is licenced mainly under the GPL version 2.0[0], and other compatible licenses, and provides a section and 2 paragraphs on how there is no warranty for those who decide to use it.

"""

NO WARRANTY

11. BECAUSE THE PROGRAM IS LICENSED FREE OF CHARGE, THERE IS NO WARRANTY FOR THE PROGRAM, TO THE EXTENT PERMITTED BY APPLICABLE LAW. EXCEPT WHEN OTHERWISE STATED IN WRITING THE COPYRIGHT HOLDERS AND/OR OTHER PARTIES PROVIDE THE PROGRAM "AS IS" WITHOUT WARRANTY OF ANY KIND, EITHER EXPRESSED OR IMPLIED, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE. THE ENTIRE RISK AS TO THE QUALITY AND PERFORMANCE OF THE PROGRAM IS WITH YOU. SHOULD THE PROGRAM PROVE DEFECTIVE, YOU ASSUME THE COST OF ALL NECESSARY SERVICING, REPAIR OR CORRECTION.

12. IN NO EVENT UNLESS REQUIRED BY APPLICABLE LAW OR AGREED TO IN WRITING WILL ANY COPYRIGHT HOLDER, OR ANY OTHER PARTY WHO MAY MODIFY AND/OR REDISTRIBUTE THE PROGRAM AS PERMITTED ABOVE, BE LIABLE TO YOU FOR DAMAGES, INCLUDING ANY GENERAL, SPECIAL, INCIDENTAL OR CONSEQUENTIAL DAMAGES ARISING OUT OF THE USE OR INABILITY TO USE THE PROGRAM (INCLUDING BUT NOT LIMITED TO LOSS OF DATA OR DATA BEING RENDERED INACCURATE OR LOSSES SUSTAINED BY YOU OR THIRD PARTIES OR A FAILURE OF THE PROGRAM TO OPERATE WITH ANY OTHER PROGRAMS), EVEN IF SUCH HOLDER OR OTHER PARTY HAS BEEN ADVISED OF THE POSSIBILITY OF SUCH DAMAGES.

"""

No Linux kernel engineer is obliged to fix any bugs, even security vulnerabilities, if one decides to, but since there's so much at stake, kernel engineers will end up fixing at their own timeframes, as they see fit.

[0] https://www.gnu.org/licenses/old-licenses/gpl-2.0.en.html

>and obviously they're not going to make any guarantees to randos

This is what I mean by purposefully misreading my comments. I have never once said or asked for this. You are arguing in bad faith.

You literally started by calling them scummy. That is an implicit demand for change.

My complaint was not about the license, it was about how the product represents itself.

Linux.org is a much different website than keycloak.org in the way they represent themselves, how they communicate the product, etc.

And you also shouldn't expect anyone to use your software. Which of course is up to you.

Not necessarily. Red Hat issues patches and backports to customers regularly and those don’t necessarily flow upstream right away (or sometimes ever).

Though Red Hat did a search and replace for RHSSO with "Red Hat Build of Keycloak" in their docs, and now they are extremely painful to read with Red Hat Build of Keycloak sometimes appearing three times in one sentence when "Keycloak" or some other shortened form would suffice.

Nobody has an obligation to even do that outside of a contractual relationship, though it would be polite.