Cacti Command Injection Vulnerability

What CISA says to do

Apply updates per vendor instructions.

Cacti contains a command injection vulnerability that allows an unauthenticated user to execute code.

• github.com
• nvd.nist.gov

What the flaw is

Cacti is an open source platform which provides a robust and extensible operational monitoring and fault management framework for users. In affected versions a command injection vulnerability allows an unauthenticated user to execute arbitrary code on a server running Cacti, if a specific data source was selected for any monitored device. The vulnerability resides in the `remote_agent.php` file. This file can be accessed without authentication. This function retrieves the IP address of the client via `get_client_addr` and resolves this IP address to the corresponding hostname via `gethostbyaddr`. After this, it is verified that an entry within the `poller` table exists, where the hostname corresponds to the resolved hostname. If such an entry was found, the function returns `true` and the client is authorized. This authorization can be bypassed due to the implementation of the `get_client_addr` function. The function is defined in the file `lib/functions.php` and checks serval `$_SERVER` variables to determine the IP address of the client. The variables beginning with `HTTP_` can be arbitrarily set by an attacker. Since there is a default entry in the `poller` table with the hostname of the server running Cacti, an attacker can bypass the authentication e.g. by providing the header `Forwarded-For: <TARGETIP>`. This way the function `get_client_addr` returns the IP address of the server running Cacti. The following call to `gethostbyaddr` will resolve this IP address to the hostname of the server, which will pass the `poller` hostname check because of the default entry. After the authorization of the `remote_agent.php` file is bypassed, an attacker can trigger different actions. One of these actions is called `polldata`. The called function `poll_for_data` retrieves a few request parameters and loads the corresponding `poller_item` entries from the database. If the `action` of a `poller_item` equals `POLLER_ACTION_SCRIPT_PHP`, the function `proc_open` is used to execute a PHP script. The attacker-controlled parameter `$poller_id` is retrieved via the function `get_nfilter_request_var`, which allows arbitrary strings. This variable is later inserted into the string passed to `proc_open`, which leads to a command injection vulnerability. By e.g. providing the `poller_id=;id` the `id` command is executed. In order to reach the vulnerable call, the attacker must provide a `host_id` and `local_data_id`, where the `action` of the corresponding `poller_item` is set to `POLLER_ACTION_SCRIPT_PHP`. Both of these ids (`host_id` and `local_data_id`) can easily be bruteforced. The only requirement is that a `poller_item` with an `POLLER_ACTION_SCRIPT_PHP` action exists. This is very likely on a productive instance because this action is added by some predefined templates like `Device - Uptime` or `Device - Polling Time`. This command injection vulnerability allows an unauthenticated user to execute arbitrary commands if a `poller_item` with the `action` type `POLLER_ACTION_SCRIPT_PHP` (`2`) is configured. The authorization bypass should be prevented by not allowing an attacker to make `get_client_addr` (file `lib/functions.php`) return an arbitrary IP address. This could be done by not honoring the `HTTP_...` `$_SERVER` variables. If these should be kept for compatibility reasons it should at least be prevented to fake the IP address of the server running Cacti. This vulnerability has been addressed in both the 1.2.x and 1.3.x release branches with `1.2.23` being the first release containing the patch.

From the NVD.

Fixes and advisories

• github.comPatch
• github.comPatch
• github.comPatch
• github.comMitigation, Patch
• github.comPatch
• github.comPatch

On SentryFeed

No story in the last 14 days has mentioned it.