Crisp's blog

Cross-site Request Forgery is a very common social exploit method to make people unknowingly do things on their own behalf on a targeted website. It's the number four on the 2010 CWE/SANS Top 25 Most Dangerous Programming Errors list.

The main reason this problem exists in most websites is the fact that they don't check the origin of an incoming request that results in an action on that website. There are several ways a website can protect itself against these sort of attacks and I'm going to explain the way we, at Tweakers.net, have implemented our own protection method.

I'm not saying that Tweakers.net is 100% safe against CSRF attacks; I know we're not, and a recent topic on our forum made that somewhat painfully clear again: with the redesign of our account pages we separated certain data into several forms, but forgot to add our token-based security measure to each of them.

Adding a special token to an HTML form is a way to prevent these CSRF attacks. Such tokens can be used to verify whether the request to change data to your website actually originated from one of your own pages and thus may be considered 'valid'. We have been using our own token-system for quite some time and in various (but not all) places on our website, but our frontpage token-system suffered from some serious drawbacks.

On our forum we are using the user's session-id as a token for forms that can be used to manipulate data. This in itself is fine, but it means that session-id's are being exposed in the mark-up data, and as a token they are not actually limited to anything (except maybe on IP-address from the originating request).

When we designed our token system for the frontpage we wanted some more security; we wanted tokens that can only be used once for each request and that were tied to a user's session-id (or IP-address in case of not-logged-in users). So we simply set up a system that created random tokens and wrote them to the database with a creation-time and the user's session-id or IP-address. Tokens older than 24 hours were removed (and thus invalidated).

This has worked well for a long time, but performance-wise became a real bottleneck: the random insertions of tokens (the random token id being the primary key of the table) in a table having always around 500.000 'active' keys sometimes slowed down page-generation enormously.

When you have a table containing over half a million tokens of which only a very small percentage will actually be used you are most certainly doing something wrong, so we decided to change that. We only wanted to store tokens that where actually used to prevent re-use, and let all other other tokens expire by itself. The question was how to accomplish that without needing storage of those tokens. The answer was: creating tokens that can validate themselves against their use and expiration

The idea is simple: take those items that you want to be validated and hash them with a secret key. The result will be the token used in the form, and server-side we can recreate that same token and compare it to the value received from the request. That takes care of almost everything, except for the expiration part. We need to be able to determine the expiration 'plainly' from the token itself. Luckily, if you add the same expiration to the items that are being hashed, it is no problem to add it in a readable format to the token. Also, we can use this expiration as a timestamp to store the tokens that have been used, together with the actual token itself to create an unique key with a primary key that is not random but actual rather sequential. We added another 6-byte random identifier in order to create randomness for the same expiration timestamps.

So in short, this is what we do:

• determine expiration timestamp
• create a random 6-byte string
• hash together expiration timestamp + an action identifier (that binds the token to a certain action or page) + session-id or IP adress using a key consisting of the random 6-byte string and a secret key
• combine above 3 values into one single token

Even though the expiration timestamp is directly readable from the combined token itself it isn't changeable because it would result in an other hash being recreated upon validation. Also by adding an 'action' identifier we were able to restrict tokens to a certain use. With the expiration being added we are more flexible in how long our tokens should be valid and can (and have) reduced that significantly compared to the 24 hours that our old token-system used.

Here's a simplified version of our token-class that illustrated this method:

PHP:

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 192 193 194 195 196 197 198 199 200 201 202 203 204 205 206 207 208 209 210 211 212 213 214 215 216 217 218 219 220 221 222 223 224 225 226 227 228 229 230 231 232 233 234 235 236

<?php /* CREATE TABLE `form_tokens` (   `token_expires` int(10) unsigned NOT NULL,   `token` binary(20) NOT NULL,   PRIMARY KEY (`token_expires`,`token`) ) ENGINE=InnoDB; */ /**  * Class to validate form-submits to prevent CSRF attacks.  *  * This allows us to check whether a user had earlier requested the form from us  * and can also be used to prevent resubmitting a form. This specific implementation  * uses "self validating" tokens, for more information, see this blogpost:  * http://crisp.tweakblogs.net/blog/3928/csrf-protection-with-self-validating-tokens.html  *  * To simply get a token to use 'somewhere', use FormToken::generateToken(),  * to get a bit of ready made html, use FormToken::generateTokenHtml().   */ class FormToken {     const defaultTtl = 7200;     const formTokenName = 'token';     const tokenKey = 'this is my little secret';          private $action;     private $token;     private $tokenError;     /**      * Contstruct a new FormToken-instance used for validating some submitted form token.      *      * @param string|bool $action The action to use for validation, use true to fall back to the current script's filename.      * @param string|bool $token The token to use for validation, use false to indicate retrieving it from the HTTP request parameters.      */     public function __construct($action=false, $token=false)     {         $this->action = $action === true ? $this->getCurrentAction() : $action;         $this->token = $token === false ? $this->getTokenFromRequest() : $token;     }     /**      * Generate a token for a specific form.      *      * This ties the generated token to the given action, so it can't be reused      * outside the specific form. Obviously leaving the action empty here and while      * validating mitigates that limitation.      *      * @param string|bool $action The form-action where this token is placed in, use true to use the current script's filename.      * @param int $ttl The time the token is valid from the script's start-up time moment on.      * @return string A base64 encoded string representing the token data.      */     public static function generateToken($action=false, $ttl=self::defaultTtl)     {         if ($action === true)             $action = self::getCurrentAction();         $timeValid = time() + $ttl;         $prefix = self::getRandomHexString(6);         $timeHex = sprintf('%08x', $timeValid);         $hmac = self::getTokenHmac($prefix, $timeValid, $action);         return base64_encode(pack('H*', $prefix . $timeHex . $hmac));     }     /**      * Generate a token wrapped in a ready-made hidden input HTML-element.      *      * @param string|bool $action The form-action where this token is placed in, use true to use the current script's filename.      * @param int $ttl The time to live.      * @return string A string representing a hidden input with a new token.       * @see generateToken()      */     public static function generateTokenHtml($action=false, $ttl=self::defaultTtl)     {         $token = self::generateToken($action, $ttl);         return '<input type="hidden" name="' . self::formTokenName . '" value="' . $token . '">';     }     /**      * Validate the representated token.      *      * This method may also be used to actually prevent reuse of the token, by storing it in the database.      *      * @param bool $storeToken Whether to store the token to prevent reuse.      * @return bool Whether the token was valid and not used previously.      */     public function validateToken($storeToken=false)     {         if (empty($this->token))         {             $this->tokenError = 'token_missing';             return false;         }         if (strlen($this->token) == 40)         {             $hex = bin2hex(base64_decode($this->token, true));             if ($hex)             {                 // Reconstruct the token's elements                 $timeValid = hexdec(substr($hex, 12, 8));                 if ($timeValid < time())                 {                     $this->tokenError = 'token_expired';                     return false;                 }                 $shortToken = substr($hex, 20);                 $hmac = self::getTokenHmac(substr($hex, 0, 12), $timeValid, $this->action);                 if ($hmac == $shortToken)                 {                     // Write token to DB to prevent re-use                     $tokenValid = !$storeToken || @mysql_query("                         INSERT INTO                             form_tokens                         (                             token_expires,                             token                         )                         VALUES                         (                             " . $timeValid . ",                             '" . mysql_real_escape_string(pack('H*', $shortToken)) . "'                         )"                     );                     if (!$tokenValid) // duplicate key                         $this->tokenError = 'token_already_used';                     return $tokenValid;                 }             }         }         $this->tokenError = 'token_invalid';         return false;     }     /**      * Get the reason why the token was considered invalid.      *      * @return string The reason.      */     public function getTokenError()     {         return $this->tokenError;     }     /**      * Generate a HMAC-SHA1 encoded string with the given prefix and some additional identifiyng elements.      *      * This mainly uses the user's sessionid or ip to prevent token-sharing.      *      * @param string $prefix A (preferrably random) prefix for the token generation.      * @param int $time The time the token will be valid.      * @param string $action The form-action to encode with the token.      * @return string The final HMAC-binary string.      */     private static function getTokenHmac($prefix, $time, $action)     {         $data = array($time);         $data[] = session_id() ? session_id() : $_SERVER['REMOTE_ADDR'];         if ($action)             $data[] = $action;         /* you might want to add user agent as well         if (!empty($_SERVER['HTTP_USER_AGENT']))             $data[] = $_SERVER['HTTP_USER_AGENT'];         */         $key = $prefix . self::tokenKey;         $data = implode('|', $data);         $blocksize = 64;         if (strlen($key) > $blocksize)             $key = sha1($key, true);         $key = str_pad($key, $blocksize, chr(0x00));         $ipad = str_repeat(chr(0x36),$blocksize);         $opad = str_repeat(chr(0x5c),$blocksize);         return sha1(($key ^ $opad) . sha1(($key ^ $ipad) . $data, true));     }     /**      * Retrieve a random hex-encoded string.      *      * @param int $byteCount The number of bytes for the return string.      * @return string A random hex-encoded string.      */     private static function getRandomHexString($byteCount)     {         $byteString = '';         while ($byteCount--)             $byteString .= sprintf('%02x', mt_rand(0, 255));         return $byteString;     }     /**      * Retrieve the current name of the PHP script file.      *      * @return string The name of the current PHP script file.      */     private static function getCurrentAction()     {         return $_SERVER['SCRIPT_NAME'];     }     /**      * Retrieve a token from the POST or otherwise GET data.      *      * @param string $tokenName The token's html-element name in the form.      * @return bool|string The token to retrieve or false if none was found.      */     private function getTokenFromRequest($tokenName=self::formTokenName)     {         if (isset($_POST[$tokenName]))             return $_POST[$tokenName];         if (isset($_GET[$tokenName]))             return $_GET[$tokenName];         return false;     } } ?>

To add a token to your HTML form, just add the output from
FormToken::generateTokenHtml([action]);
somewhere within the form itself, where [action] is an identifier to identify the current action (you can use the boolean 'true' to fall back to the current script's name). To validate, you can do something like this:

PHP:

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26

<?php $formToken = new FormToken([action]); if (!$formToken->validateToken(true)) {     switch ($formToken->getTokenError())     {         case 'token_missing':             echo 'There was no token in the request.';         break;         case 'token_expired':             echo 'The token was expired.';         break;         case 'token_already_used':             echo 'The token was already used before.';         break;         case 'token_invalid':         default:             echo 'The token was invalid.';     } } else {     // proceed } ?>

You may use this code in any way you like, it's not licensed whatsoever. Please help make the web a safer place!