グローバル変数の登録機能の使用法
@f dot zijlstra
Actually you should not filter against known bad data, you need to white list your filters. Rather than looking for bad data, you should strip everything but what you expect because it's future proof.
example for a 32 character text input field:
$inputvar=substr(htmlentities($_POST['inputvar']),0,32);
$inputvar=preg_replace('/[^\w\.\-\& ]/', '', $inputvar);
This will strip numbers, newlines, null characters, # and all non-printable characters, only allowing word characters(determined by locale setting), .,-, space and &, as well as truncate the value to the size you accept. You could use preg_match instead and throw an error if there's anything outside the pattern.
This can be a little painful because you might miss a character you need, and have to add it to your code, but it's easier than restoring your database, or explaining to your customers about why their data is now in the hands of identity thieves.
Default deny for the long term win. It's the only security panacea that works now and always will. Blacklists are the completely wrong approach because they don't handle future threats you don't know about yet.
Further you should use mysqli and prepared statements for queries which use user data as parameters, after using the mysql filters on the input data. If you really want to go the extra mile, start using stored procedures because it will allow you to remove SELECT, DELETE, UPDATE and INSERT permissions from the web server user. All you need is EXECUTE for stored procs, thereby limiting what can be seen of the database to only the functions you expose(and you can easily expose too much if you aren't careful).
Naturally the big dangers are SQL injection and XSS. A simple filter like this will break them when combined with a prepared statement.
As someone else noted, relying on _anything_ from the client is an extremely bad idea because it can all be spoofed (including HTTP_REFERER) using Paros, Tamper Data or any other client side proxy, by doing a simple man in the middle tamper on yourself.
You can make HTTP_REFERER=http://www.whitehouse.gov if you want.
Sorry to segway into db security, but you can't avoid it when talking about sanitizing input which leads to building queries.
-Viz
ユーザが投稿したデータ
多くの PHP のプログラムで最も脆弱な部分は、言語自体に起因するものではなく、 単にセキュリティを考慮して書かれていないコードの問題です。そのため、 指定したコードの部分の意味を常に時間をかけて吟味し、 予想外の変数が投稿された場合に有り得る損害を確かめる必要があります。
例1 危険な変数の使用
<?php
// ユーザのホームディレクトリからファイルを削除します... または他の誰
// かのディレクトリかも?
unlink ($evil_var);
// 彼らのアクセスのログを書き込む.. または違うかも?
fputs ($fp, $evil_var);
// 何かちょっとしたことを実行.. または rm -rf *?
system ($evil_var);
exec ($evil_var);
?>
- このスクリプトは、意図したファイルのみを受け付けるか?
- 例外的なまたは意図したもの以外のデータにより実行することが可能 か?
- このスクリプトは意図した以外の方法で使用することが可能か?
- このスクリプトは、悪い意味で他のスクリプトと組み合わせて使用す ることが可能か?
- トランザクションは適切に記録されているか?
register_globals,magic_quotes, または他の便利な設定は、有効性、発 信元、指定した変数の値について混乱を生じる可能性があるため、設定を オフにしたいと思うかもしれません。error_reporting(E_ALL) モードで PHPを動作させた場合、確認または初期化する前に使用された変数に関し て警告を発生させることも可能です。(これにより、処理時に通常とは異 なるデータを防止することが可能です)
add a note
User Contributed Notesユーザが投稿したデータ
Viz
12-Jun-2008 07:47
12-Jun-2008 07:47
f dot zijlstra at gmail dot com
18-May-2008 03:07
18-May-2008 03:07
I'd like to note that the 'easysecure' thing posted below is NOT a secure way to validate that the form was indeed submitted from a browser. In fact, there is NO way you can guarantee that.
A smart person with bad intentions can easily parse the HTML page to fetch the generated token and pass it on. The only way to secure your forms is to explicitly check every variable against potentially dangerous values.
Livingstone@stonyhills[dot]com
02-Feb-2008 01:51
02-Feb-2008 01:51
making sure your form is submitted from your page! Could also be adapted to url, by additing &token to the query string and checking this against session data(or what ever array you like) with $_GET, not that this string is randomly generated and stored. If you like you could build your own array to store the generated string if you dont want to use $_SESSION, say you could make yours like $tokens = array(), and in your easysecure class you store all the stuff in that array!
<?php
class easysecure {
var $curr_user;
var $curr_permission;
var $curr_task;
var $validpermission;
var $error;
function &setVar( $name, $value=null ) {
if (!is_null( $value )) {
$this->$name = $value;
}
return $this->$name;
}
function maketoken($formname, $id){
$token = md5(uniqid(rand(), true));
$_SESSION[$formname.$id] = $token;
return $token;
}
function checktoken($token, $formname, $id){
//print_r($_SESSION);
//echo ($token);
//if we dont have a valid token, return invalid;
if(!$token){
$this->setVar('validpermission', 0);
$this->setVar('error', 'no token found, security bridgedetected');
return false;
}
//if we have a valid token check that is is valid
$key = $_SESSION[$formname.$id];
if($key !== $token ){
$this->setVar('validpermission', 0);
$this->setVar('error', 'invalid token');
return false;
}
if($this->validpermission !==1){
echo 'invalid Permissions to run this script';
return false;
}else{
return true;
}
}
}
?>
<?php $userid = *** //make it what ever id you like ?>
<form name="newform" action="index.php" method="post">
<input type="text" name="potentialeveilfield" value="" size 30 />
<input type="hidden" name="token" value="<?php echo maketoken(newform, $userid); //$userid here could be user profile id ?>" />
<input type="submit" />
</form>
Now when processing the form... check the value of your token
<?php
//well you know the form name
if(!checktoken($_POST['token'], 'newform', $userid))
{ //failed
exit(); //or what ever termination and notification method best suits you.
//you could also design the class your way to get more accurate fail (error messages from the var)
}
//you can now continue with input data clean up (validation)
?>
Uli Kusterer
13-Sep-2005 12:50
13-Sep-2005 12:50
One thing I would repeat in the docs here is what information actually comes from the user. Many people think a Cookie, since it's written by PHP, was safe. But the fact is that it's stored on the user's computer, transferred by the user's browser, and thus very easy to manipulate.
So, it'd be handy to mention here again that:
CGI parameters in the URL, HTTP POST data and cookie variables are considered "user data" and thus need to be validated. Session data and SQL database contents only need to be validated if they came from untrustworthy sources (like the ones just mentioned).
Not new, but I would have expected this info under this headline, at least as a short recap plus linlk to the actual docs.