If you're running any PHP based website on NGINX server and have PHP-FPM feature enabled for better performance, then beware of a newly disclosed vulnerability that could allow unauthorized attackers to hack your website server remotely.
The vulnerability, tracked as CVE-2019-11043, affects websites with certain configurations of PHP-FPM that is reportedly not uncommon in the wild and could be exploited easily as a proof-of-concept (PoC) exploit for the flaw has already been released publicly.
PHP-FPM is an alternative PHP FastCGI implementation that offers advanced and highly-efficient processing for scripts written in PHP programming language.
The main vulnerability is an "env_path_info" underflow memory corruption issue in the PHP-FPM module, and chaining it together with other issues could allow attackers to remotely execute arbitrary code on vulnerable web servers.
The vulnerability was spotted by Andrew Danau, a security researcher at Wallarm while hunting for bugs in a Capture The Flag competition, which was then weaponized by two of his fellow researchers, Omar Ganiev and Emil Lerner, to develop a fully working remote code execution exploit.
Which PHP-based websites are vulnerable to hackers?
Though the publicly released PoC exploit is designed to specifically target vulnerable servers running PHP 7+ versions, the PHP-FPM underflow bug also affects earlier PHP versions and could be weaponized in a different way.
In brief, a website is vulnerable, if:
- NGINX is configured to forward PHP pages requests to PHP-FPM processor,
- fastcgi_split_path_info directive is present in the configuration and includes a regular expression beginning with a '^' symbol and ending with a '$' symbol,
- PATH_INFO variable is defined with fastcgi_param directive,
- There are no checks like try_files $uri =404 or if (-f $uri) to determine whether a file exists or not.
This vulnerable NGINX and PHP-FPM configuration looks like the following example:
Here, the fastcgi_split_path_info directive is used to split the URL of PHP web pages into two parts, the value of one help PHP-FPM engine to learn the script name and the other one contains its path info.
How does PoC RCE exploit for PHP FPM work?
According to the researchers, the sample regular expression, which defines the fastcgi_split_path_info directive, as shown, can be manipulated by using the newline character in a way that the split function eventually sets the path info empty.
Next, since there is an arithmetic pointer in FPM code that incorrectly assumes that env_path_info has a prefix equal to the path to the php script without actually verifying the existence of the file on the server, the issue can be exploited by an attacker to overwrite data in the memory by requesting specially crafted URLs of the targeted websites.
In the background, the PoC exploit [1 (PHuiP-FPizdaM), 2 ] researchers released chains together both these issues to manipulate the memory and add custom php.ini values, as shown in the screenshot, in the PHP-FPM configuration file of a targeted server, allowing attackers to execute arbitrary code using a web-shell.
"Using a carefully chosen length of the URL path and query string, an attacker can make path_info point precisely to the first byte of _fcgi_data_seg structure. Putting zero into it moves `char* pos` field backward, and following FCGI_PUTENV overwrites some data (including other fast cgi variables) with the script path," researchers said in a bug report submitted to the PHP project.
"Using this technique, I was able to create a fake PHP_VALUE fcgi variable and then use a chain of carefully chosen config values to get code execution."
PHP 7 updates released to patch FPM flaw
The list of preconditions for successful exploitation, as mentioned above, is not uncommon because the vulnerable configurations are being used by some of the web hosting providers and available on the Internet as part of many PHP FPM tutorials.
One such affected web hosting provider is Nextcloud who released an advisory yesterday warning its users that "the default Nextcloud NGINX configuration is also vulnerable to this attack" and recommending system administrators to take immediate actions.
A Patch for this vulnerability was released just yesterday, almost a month after researchers reported it to the PHP developer team.
Since the PoC exploit is already available and the patch released just yesterday, it's likely possible that hackers might have already started scanning the Internet in search for vulnerable websites.