Datasets:
Devilishcode
/
hacking

Dataset card Data Studio Files
xet
 Community
text
stringlengths
0
1.99k
the metadata itself is also stored in a *serialized format* - which, as it
turns out, opened a whole new door for attackers.
So, how can we exploit this serialized field? Let's reuse the "Blind AFR"
from earlier, but this time we'll change the function from "reading a
file" to something even more restricted - checking if a file exists:
=> Try harder: Blind Arbitrary File-Check
---------------------------------------------------------------------------
<?php file_exists( $_GET['file'] );
---------------------------------------------------------------------------
At first glance, it might seem like filter chains could help again.
However, since `file_exists()` literally only checks if a file exists
without actually processing its content, you can't apply the previous
side-channel oracle here. But here's another twist - in order to
conveniently use PHAR files within PHP scripts, PHP introduced the
`phar://` built-in wrapper back in PHP 5.3. And whenever PHP parses a PHAR
file with this protocol, it automatically deserializes the metadata stored
inside. This means nearly every file operation in PHP could potentially
become another entry point for deserialization!
As for exactly how we can escalate this from PHAR deserialization all the
way to remote code execution, there are still some practical challenges to
overcome - such as figuring out how to deliver a malicious PHAR file onto
the remote server. (Perhaps our efforts in the LFI Arms Race weren't
wasted after all!) This heavily requires the attacker's creativity and
their familiarity with the target environment. I believe Thomas already
showed an impressive RCE in TCPDF during his talk. Here, I'd like to
introduce another brilliant case involving mPDF!
==============================================
| Level Max - Laravel (w/ mPDF) Kill Chain |
==============================================
Just like TCPDF, mPDF is another widely used library when you need to
convert HTML into PDFs. And during the conversion process, mPDF performs
file operations on image URLs as well - meaning attackers can easily reuse
the same technique to trigger PHAR deserialization:
----------------------------[ So PHAR so Good! ]---------------------------
<img src="phar://path/to/image.jpg" />
---------------------------------------------------------------------------
The issue was first discovered [76] back in 2019 and promptly got patched.
However, @Cyku quickly found another way to trigger the vulnerability and
provided a full exploit [77] based on a real-world scenario! He also
discovered that mPDF actually caches embedded Data URIs onto the remote
filesystem. By exploiting the predictable randomness of the cached
filenames, he was able to smuggle a crafted PHAR file - then combine it
with Laravel's built-in POP chains - to finally achieve RCE!
------------------------[ Exploit mPDF All in One! ]-----------------------
<style>
background: url(data:image/jpeg;base64,HERE-IS-PHAR-PAYLOAD-IN-BASE64);
</style>
<img src="#" ORIG_SRC="phar:///tmp/mpdf/_tempCSSidata42_0.jpeg/a.jpg"/>
---------------------------------------------------------------------------
The PHAR mechanism really opened up a whole new era of PHP
deserialization. As more researchers got involved, this attack surface
gradually expanded to cover more applications, libraries, and even PHP
frameworks. Ultimately, this forced the PHP team to disable automatic
deserialization in the PHAR protocol starting from PHP 8.0 [78]. I'm sure
that was fantastic news for both Thomas and me - because it meant that our
"security research" actually did something positive in the real world, and
made PHP a little bit safer! :)
----[ 5. Participants Also Popped 0days
When we talk about "an awesome CTF," I'm not sure which name immediately
pops into your mind. In my opinion, while high-quality challenges and
experienced organizers are important, it's the participants themselves who
truly make a CTF awesome.
I believe we've already shown how the CTF community works: participants
can not only learn new tricks straight from challenge authors, but authors
themselves can also discover their own blind spots through unintended
solutions. Both sides push each other forward, working together to advance
the entire Infosec community!
But sometimes, this kind of interaction can get a bit "out of hand." We've
seen plenty of cases where the unintended solutions submitted by CTF
players turned out to be actual 0days - that happened repeatedly in
Chromium [79], VirtualBox [80], and even CS:GO [81]. Sometimes, even the
CTF authors expect players to solve the challenges using unknown 0days. As
far as I know, certain CTFs also have a special "Zajebiste" category,
specifically for these challenges involving 0days or something very close
to it!