Steal My Bitcoin CTF PoC
github.comTo be fair, the person who created the challenge stated that is was a fairly standard setup, replicating something outside a lab, but in reality is was something completely different, and not anywhere near how PHP is usually used.
It used the development server in PHP, and did not even run requests with fpm or mod_php. Instead, it used the built in dev server of PHP, and just proxied requests from nginx to the php dev server.
Literally the first thing on top of the PHP docs about the dev server used here is this: "Warning This web server is designed to aid application development. It may also be useful for testing purposes or for application demonstrations that are run in controlled environments. It is not intended to be a full-featured web server. It should not be used on a public network."
That is not something people do, and it also explains why the site was down all the time, cause the dev server is super slow, as it's not meant for this in any way shape or form. A real setup would add a few zeros to the requests per second a server could handle.
So all in all, this was very pointless, as it was nothing like real php sites. A shame, cause a realistic setup woild be more interesting and relevant to answer the question of how hard is it really to gain RCE on a normal site.
> "It is not intended to be a full-featured web server. It should not be used on a public network."
Behind nginx or similar isn't exactly on a public network. I don't think I'd run it that way, but I could see someone else doing it if they read the docs a certain way.
That's interesting because there were people complaining yesterday that this was a fully static website with no PHP involved lol
https://news.ycombinator.com/item?id=50234254
OP confirmed that the server was initially serving only an index.html. In that configuration, the vulnerable code path is never reached — it requires a .php file. OP added php-version.php roughly 18–19 hours into the experiment (I'm not sure of the exact timing), so that period should be subtracted from the publicly displayed challenge duration.
HITCON CTF has a realistic php2shell challenge. Think it has ended up with 0 solves so far, not sure.
The "realistic setup" used for this didn't have a php page on it initially.
I'm skeptical of the value of such setups, unless it's clear it's an out of the box setup, and we already know those are totally broken.
I don't understand who these people think they are really fooling. Is this some AI denial?
There are many PoCs online that work on realistic setups (like PHP docker image). If you're curious you can try to replicate it locally.
there are a number of php docker images. some are insecure by default.
many php setups sit behind nginx, caddy, or other webserver with a max body size limit.
anyone following OWASP / similar guidelines was probably okay here (for this particular PoC path), if not using apache/mod_php, i would think.
are you saying this lead to dos (and less competition)? or easier penetration?
At first glance, this wouldn't affect demonstration of the vuln.
Related: https://x.com/JamesPavur/status/2114501212313653696
Apparently the wallet key wasn't on the server, like some in the original thread suspected, but perhaps not due to maliciousness after all.
This is already linked by it's author in the Steal My Bitcoin thread (585 points, 458 comments) https://news.ycombinator.com/item?id=50224916
between the challenge being updated to actually run php and this only like 2-3 hours passed
Does php2shell works with PHP-fpm nginx? I can’t truly replicate that poc locally.
in theory.
many common nginx setups and hardening suggestions would make it very difficult to exploit. as well as any sort of distributed architecture that doesn't pin requests to clients.
there could be other undiscovered paths to exploit this vulnerability though. the PoC here uses the disclosed path for exploiting on patch day, minus mod_php.
I believe this managed to prove something.
Namely that the vulnerability was not that easy to exploit, meaning that the myriads of servers that stayed unpatched during those initial days are probably not all members of botnets now. (Contrary to what the discourse might have made it seem like.)
It took about 2 hours to get taken over and that was with a setup that's sufficiently non-standard that existing PoC would've needed tweaking.
The fact that people are still in denial over this is crazy.
I don't understand what the upside was supposed to be from denying it. This has been an exceptionally weird couple of days. Was it an anti-AI thing?
It’s been going on for far longer, people want to believe all OpenAI and Anthropic are doing is pure marketing. When the first Fable results were released, everybody was skeptical (which is rational), but with every third party confirmation of the capabilities they doubled down that this must be even more marketing instead of admitting that they were wrong. To the point that they are no longer skeptical but in pure denial about what is happening.
I suspect so, and just a general blowback effect where if what people perceived to be the message is even slightly more dramatic than the reality then there's tendency for many to overshoot reality in the other direction.
For me the key question is whether tens of millions of random unimportant websites were mass-hacked by AI agents creating sleeper botnets.
If you think so then express it clearly, I'm interested in those thoughts.
–––
Apparently the challenge was dubious so I’m not sure what to draw from it in the end.
A bunch of folks tried to say the challenge was dubious at best and that the vulnerability was a legitimate concern, and were told by commenters here that they were fearmongering or shills.
Evidence has shown they were pretty spot on about the first part (the test originally served static html, the php code wasn’t actually as vanilla as described, and the bitcoin key wasn’t actually there).
I’d have hoped this kind of clear vindication on the first point might have caused folks to revisit their stance on whether to trust the expert stance on the second. But maybe I’ve been overly optimistic.
Just outrage/security fatigue I believe.
Where do you get the two hours from? Did this attacker spend two hours on it, after 17 hours of others failing to break in?
> The fact that people are still in denial over this is crazy.
Maybe you misunderstand what I say.
Imagine the tens of millions of personal or small business websites running php. What's your guess, what portion of the unpatched ones is already pwned? Is it 1%? 5%? 15%? 50%? 95%?
What would you guess?
According to the tweet below, the server didn't actually run any PHP code for the first 17 hours.
I would take the opposite conclusion: the initial exploiting took _only_ 19 hours due the non-standard setup, but now the setup can be trivially exploited by tuning the PoC. By now, bad actors will probably have tools that can exploit most common setups.
So what portion of the unpatched servers is pwned now? What's your guess?
I was sick when the news about php2shell broke, so ended up leaving two of my personal sites that use php un-patched for about 48 hours until my fever broke and I patched them. I didn’t get hacked. Was I just lucky and so shouldn’t infer anything about how hard it is to actually make use of this vulnerability? Probably.
Another theory is your server got compromised but they cleared their tracks. Not every hacker wants to make a quick few cents from a miner that gets shut down in a day or two.
Apparently this could only be solved after the challenge added a php file that shows the PID. Which is not public information.
echo 'PID: ' . getmypid();
I used Kimi K3.1 Pro to figure it out as soon as cloudlandsdev put php-version.php live. The exploit worked just like this in a local Docker lab. But the terms page of the contest prohibited "resource exhaustion," so it stopped there. Showing Kimi the PoC, it says:
"The winning chain deliberately used per-request memory exhaustion, which appears to violate the literal no-resource-exhaustion rule, although the organizer accepted the win."
Yeah, each agent stumbled upon this. While one of those CVS IS memory exhaustion.
it doesn't sound like you understand what you are doing?
Why? He clearly explained it, rules prohibited resource exhaustion literally, which is funny because it prohibited to do the exploit it was asked.
This was good wholesome fun! Some foundation should do this for future cve's
I actually had GLM 5.3 run against the site on the day, but it was around 2 am and I was pretty sleepy by 3. GLM found the memory overflow vulnerability but I stopped it before it could continue into retrieving something useful from it.
In retrospect, I should have just given it a /goal and let it run overnight.
As an outsider this was fun to watch. Namely, the dozens of l33t haxors including so-called "professionals" who claimed they'd be able to solve this in a jiffy with their magic exploits, yet couldn't.
The OP shows that it was, in fact, solved in a jiffy (a few hours work done largely by an LLM) with a magic (generally applicable) exploit. Given that, I suspect the reputable security experts in the thread who claimed they could solve it were telling the truth.
The initial server setup wasn't realistic and it wasn't exploitable apparently.
The only professional I saw making that claim was the CEO of trailofbits who is a former NSA TAO member and doesn’t have anything to prove by talking a big game.
It wasn’t that difficult of an exploit, to be honest. I myself am not the leetest of the leet but I was most of the way there, having been set back by the downtime and my time zone. It’s definitely a fiddly, tricky exploit as most memory exploits are but I think the initial alarm was justified.
While waiting for the server to come back up, it took me about 3 hours with LLM assistance to write a php2shell RCE for nginx and mediawiki on the same version of php as TFA.
The biggest win, as others have also observe, was the demonstration that there aren’t swarms of automated pwnbots out there that will insta-root your server.
Considering the amount of downtime, I think that flag got got pretty quickly.
> there aren’t swarms of automated pwnbots out there that will insta-root your server
There are, they just didn't have a working exploit. Anyone with a honeypot can tell you attempts were widespread going back to early last week, but it wasn't until Saturday that we started to see successes.
> to write a php2shell RCE for nginx and mediawiki on the same version of php as TFA.
Ok but a mediawiki page is obviously not comparable
People were complaining that originally it was just using php to serve a static index.html file
It looks like after 18-19hours they updated it to serve a real .php file and it was hacked less than 2 hours after that