Understanding SQL Injection Basics

Introduction

Structured Query Language (SQL) is used to query, operate, and administer database systems such as Microsoft SQL Server, Oracle, or MySQL. The general use of SQL is consistent across all database systems that support it; however, there are intricacies that are particular to each system.
Database systems are commonly used to provide backend functionality to many types of web applications. In support of web applications, user-supplied data is often used to dynamically build SQL statements that interact directly with a database. A SQL injection attack is an attack that is aimed at subverting the original intent of the application by submitting attacker-supplied SQL statements directly to the backend database. Depending on the web application, and how it processes the attacker-supplied data prior to building a SQL statement, a successful SQL injection attack can have far-reaching implications. The possible security ramifications range from authentication bypass to information disclosure to enabling the distribution of malicious code to application users.
This white paper will describe SQL injection attacks, how they are performed, and precautions that should be taken inside applications or networks to reduce risks that are associated with SQL injection attacks.

SQL Injection Explained

A SQL injection attack involves the alteration of SQL statements that are used within a web application through the use of attacker-supplied data. Insufficient input validation and improper construction of SQL statements in web applications can expose them to SQL injection attacks. SQL injection is such a prevalent and potentially destructive attack that the Open Web Application Security Project (OWASP) lists it as the number one threat to web applications.

Ramifications of Successful SQL Injection Attacks

Although the effects of a successful SQL injection attack vary based on the targeted application and how that application processes user-supplied data, SQL injection can generally be used to perform the following types of attacks:

 



Authentication Bypass: This attack allows an attacker to log on to an application, potentially with administrative privileges, without supplying a valid username and password.

Information Disclosure: This attack allows an attacker to obtain, either directly or indirectly, sensitive information in a database.

Compromised Data Integrity: This attack involves the alteration of the contents of a database. An attacker could use this attack to deface a web page or more likely to insert malicious content into otherwise innocuous web pages. This technique has been demonstrated via the attacks that are described in Mass exploits with SQL Injection at the SANS Internet Storm Center.

Compromised Availability of Data: This attack allows an attacker to delete information with the intent to cause harm or delete log or audit information in a database.

Remote Command Execution: Performing command execution through a database can allow an attacker to compromise the host operating system. These attacks often leverage an existing, predefined stored procedure for host operating system command execution. The most recognized variety of this attack uses the xp_cmdshell stored procedure that is common to Microsoft SQL Server installations or leverages the ability to create an external procedure call on Oracle databases.

An Example of SQL Injection for Authentication Bypass

One of the many possible uses for SQL injection involves bypassing an application login process. The following example illustrates the general operation of a SQL injection attack. The following HTML form solicits login information from an application user. Although this example uses an HTTP POST request, an attacker could also use HTML forms that use the HTTP GET method.

<form action="/cgi-bin/login" method=post>

 Username: <input type=text name=username> 

 Password: <input type=password name=password> 

<input type=submit value=Login>

When a user enters his or her information into this form and clicksLogin, the browser submits a string to the web server that contains the user's credentials. This string appears in the body of the HTTP or HTTPS POST request as:

username=submittedUser&password=submittedPassword

   

An application with a vulnerable login process may accept the submitted information and use it as part of the following SQL statement, which locates a user profile that contains the submitted username and password:

select * from Users where (username = 'submittedUser' and password = 'submittedPassword');

   

Unless an application uses strict input validation, it may be vulnerable to a SQL injection attack. For example, if an application accepts and processes user-supplied data without any validation, an attacker could submit a maliciously crafted username and password. Consider the following string sent by an attacker:

username=admin%27%29+--+&password=+

Once this string is received and URL-decoded, the application will attempt to build a SQL statement using a username of admin') -- and a password that consists of a single space. Placing these items into the previous SQL statement yields:

select * from Users where (username = 'admin') -- and password = ' ');

        

As the previous example demonstrates, the attacker-crafted username changes the logic of the SQL statement to effectively remove the password check. In the above example, an attacker could successfully log in to the application using the admin account without knowledge of the password to that account. 
The string of two dash characters (--) that appears in the crafted input is very important; it indicates to the database server that the remaining characters in the SQL statement are a comment and should be ignored. This capability is one of the most important tools that is available to an attacker and without it, it would be difficult to ensure that the malicious SQL statements were syntactically correct. 
Although the crafted field, which is the username field in the previous example, must be tailored to the vulnerable application, a large set of documented strings that are readily available on the Internet have proven successful at enabling SQL injection attacks. The previous example may be simplistic, but it illustrates the effectiveness of SQL injection attack techniques.

Blind and Second Order SQL Injection

In situations where data from a backend SQL database is not returned directly to the user or attacker, it may be necessary for an attacker to use the Blind SQL Injection technique. With this technique, an attacker can determine whether a SQL statement was executed using means other than the direct presentation of data. Using Blind SQL Injection, an attacker could perform reconnaissance, obtain sensitive information, or alter database contents, including authentication credentials.
One example of the Blind SQL Injection technique is the introduction of a delay as part of a malicious SQL statement. Depending on the database software in use, an attacker could build a SQL statement that is designed to cause a database server to perform a time-consuming action. With the MySQL database software, it may be possible to craft a SQL statement using the sleep() function. For example, incorporating sleep(10) into a malicious query will create a 10-second delay. An attacker could induce a recognizable delay on database servers that do not contain the sleep()function by executing an operating system command or time-consuming sub-query or attempting to establish an outbound HTTP connection. Should the time-consuming SQL statement be executed, the web application may take noticeably longer to respond than is typical. This method allows attackers to determine whether their SQL statements are being executed with some level of certainty.
Second Order SQL Injection attacks involve user-submitted data that is first stored in the database, then retrieved and used as part of a vulnerable SQL statement. This class of vulnerability is more difficult to locate and exploit, but Second Order SQL Injection attacks justify data validation prior to the execution of all SQL statements in an application, as well as the comprehensive use of parameterized queries.

Defending Against SQL Injection Attacks

A SQL injection attack can be detected and potentially blocked at two locations in an application traffic flow: in the application and in the network. 

Defenses in the Application

There are several ways in which an application can defend against SQL injection attacks. The primary approaches include validation of user-supplied data, in the form of whitelisting or blacklisting, and the construction of SQL statements such that user-supplied data cannot influence the logic of the statement.

Blacklisting and Whitelisting

Within an application itself, there are two approaches to input validation that can defend against SQL injection attacks: blacklisting and whitelisting. With blacklisting, specific, known malicious characters are removed from or replaced in user input. Although this approach is often implemented, largely due to the ease at which it can be accomplished, it is not effective when compared to whitelisting. Blacklisting can fail to properly handle complex obfuscation, which could allow an attacker to subvert filters and potentially inject SQL statements. This failure often occurs as a result of evolving attack techniques and filters that are not comprehensive or not implemented correctly.
Alternately, whitelisting examines each piece of user input against a list of permitted characters. This approach is more effective in mitigating the risk of SQL injection, as it is more restrictive concerning which types of input are allowed. Well-implemented whitelisting should examine each piece of user-supplied data against the expected data format.
Regardless of the approach, it will most likely be necessary to tailor the allowed characters by the input field type or by the input field class (for example, text or numeric data). The input validation and sanitization functions that are used to filter SQL injection enabling characters can potentially be generalized and used to filter characters that are indicative of cross-site scripting attacks. More information about cross-site scripting is available .Understanding Cross-Site Scripting (XSS) Threat Vectors.
The application must act deterministically when it receives invalid characters from a user. Different levels of response may be appropriate based on the circumstances in which the unexpected data is being processed and the effect it could have. For example, applications should patently reject users that submit two dash characters (--) or a semi-colon character (;) as part of the login name or password, and a high-severity alert should be sent to application administrators. However, this somewhat harsh response may not be appropriate when an application receives a single apostrophe character as part of a street address that is submitted by an authenticated user who is entering package shipping information. Nonetheless, the application should behave as intended, and a notification should be sent to application administrators if the submitted apostrophe was not handled properly by the application.
The implementation of input validation and sanitization should contain alerting functionality. This functionality should alert an application administration or development team when unexpected input has been received because it could indicate that the filters may be incorrectly blocking valid, but unexpected data or the application may be under attack. In either case, changes may be required to the filtering functionality of the application.
At first glance, it might appear that these data validation and sanitization techniques could be implemented in client-side JavaScript to be run in a user's browser. From a security perspective, however, it must be assumed that any and all types of data will originate from a user's browser, regardless of the client-imposed safeguards that are in place. Nonetheless, client-side data validation techniques can enhance application usability.

Fortifying SQL Statements

Parameterized queries in ASP.NET, prepared statements in Java, or similar techniques in other languages should be used comprehensively in addition to strict input validation. Each of these techniques performs all required escaping of dangerous characters before the SQL statement is passed to the underlying database system. The following example depicts the use of prepared statements in Java and illustrates how SQL statements are built without user-supplied data and then augmented with the data in such a manner that the structure and intent of the SQL statement cannot be altered:

String sql = "select * from Users where (username = ? and password = ?)"; 



PreparedStatement preparedStmt = connection.prepareStatement(sql); 

preparedStmt.setString(1, submittedUsername); 

preparedStmt.setString(2, submittedPassword); 



results = preparedStmt.executeQuery(); 

Note that prepared statements and similar technologies are not a panacea; used incorrectly without bind variables, they are no more secure than traditionally constructed dynamic queries. The following example illustrates the incorrect use of prepared statements in Java:

String sql = "select * from Users where (username = '" + submittedUsername + 
         "' and password = '" + submittedPassword + "')";



PreparedStatement preparedStmt = connection.prepareStatement(sql); 



results = preparedStmt.executeQuery();

In addition to ensuring that the intent of SQL statements cannot be altered by user-supplied data, applications should also catch and remove all SQL-generated error messages before they reach an end user. Although this safeguard may hinder a developer's ability to troubleshooting an application error, which can easily be overcome with additional backend logging, the presentation of SQL errors will greatly aid attackers in successfully exploiting a SQL injection vulnerability. The following example is an overly verbose error message:

com.mysql.jdbc.exceptions.MySQLSyntaxErrorException: Table 'sqlInjectionTest.test' doesn't exist 

 at com.mysql.jdbc.SQLError.createSQLException(SQLError.java:936) 

 at com.mysql.jdbc.MysqlIO.checkErrorPacket(MysqlIO.java:2985) 

 at com.mysql.jdbc.MysqlIO.sendCommand(MysqlIO.java:1631) 

 at com.mysql.jdbc.MysqlIO.sqlQueryDirect(MysqlIO.java:1723) 

 at com.mysql.jdbc.Connection.execSQL(Connection.java:3277) 

 at com.mysql.jdbc.Connection.execSQL(Connection.java:3206) 

 at com.mysql.jdbc.Statement.executeQuery(Statement.java:1232) 

 at sqlInjectionBefore.main(before.java:28) 

This error message discloses that the application is using the Java programming language and the MySQL database platform and that the queried database is named "sqlInjectionTest." Each piece of information could assist attackers in crafting their application input, which increases the odds that they will be successful.