# CVE-2023-33381-MitraStar-GPT-2741GNAC

### CVE-2023-33381: OS command injection on MitraStar GPT-2741GNAC
### Device: GPT-2741GNAC
### Firmware Version: AR_g5.8_110WVN0b7_2
### Vulnerability Description:

Upon logging in via SSH, I immediately noticed the limited set of available commands and options. It became evident that I was indeed confined to a restricted shell environment.

![image](https://github.com/duality084/CVE-2023-33381-MitraStar-GPT-2741GNAC/assets/7117259/b5840811-334d-45d5-b3e3-7863969165a3)

After exploring several known vulnerabilities, I decided to delve into the search for new ones. To begin, I logged into the administrative portal and began testing various functionalities that could potentially result in OS command injection. Within the Diagnostic menu, I came across a particular feature that allowed me to test connectivity using the ping and traceroute commands.
![image](https://github.com/duality084/CVE-2023-33381-MitraStar-GPT-2741GNAC/assets/7117259/8ab3f70a-2291-4491-a989-9c49b5c69592)

![image](https://github.com/duality084/CVE-2023-33381-MitraStar-GPT-2741GNAC/assets/7117259/ab39a419-e528-4fa4-ae4f-0c4379c5c316)

So, I decided to try something sneaky by adding a ";" character to my command. I executed "cat /etc/passwd" and guess what? The command ran successfully.

![image](https://github.com/duality084/CVE-2023-33381-MitraStar-GPT-2741GNAC/assets/7117259/d9c780f9-649e-49a7-bb19-305abff583db)

I took it a step further and executed the "sed" command to replace the restricted shell entry in the "/etc/passwd" file with a full interactive shell.

![image](https://github.com/duality084/CVE-2023-33381-MitraStar-GPT-2741GNAC/assets/7117259/729a7bf7-118b-4146-a2ec-3de32de9487e)

As clearly illustrated in the image below, the modifications I made have enabled me to login and freely execute commands like "uname" and "cat"

![image](https://github.com/duality084/CVE-2023-33381-MitraStar-GPT-2741GNAC/assets/7117259/8ee5c6ba-b4c6-4a3f-829a-ebeb5945f18b)

### Root of cause
Since I had complete admin privileges, I couldn't resist delving deeper to uncover the vulnerable component. While inspecting the browser requests, I noticed something interesting. There were two CGI files, "ping.cgi" and "DiagGeneral.cgi", being targeted by the requests.

![image](https://github.com/duality084/CVE-2023-33381-MitraStar-GPT-2741GNAC/assets/7117259/4c9bd5c9-c71b-4d04-b0b1-718195067902)

After grabbing the files, I had some fun with reverse engineering using Ghidra. Here's the interesting part: in the "ping.cgi" file, I noticed that the **PingIPAddr** parameter was being directly taken from user input without any proper sanitization. The **PingIPAddr** parameter grabbed from user input was stored for future use utilizing the **tcapi_set** function.
![image](https://github.com/duality084/CVE-2023-33381-MitraStar-GPT-2741GNAC/assets/7117259/1ce7c0ff-8964-4934-a5a2-1cf96990c20b)

Lastly, in the DiagGeneral.cgi file the PingIPAddr parameter was retrieved using the **tcapi_get** function and then directly used in the **system** function without any sanitization. This flaw creates a command injection vulnerability, enabling unauthorized execution of arbitrary commands on the system.

![image](https://github.com/duality084/CVE-2023-33381-MitraStar-GPT-2741GNAC/assets/7117259/a02e6faf-adc3-4da9-9efa-abc58f601a4f)