All relational databases—including SQL Server, Oracle, IBM DB2, and MySQL—are susceptible to SQL-injection attacks. You can buy products that protect your system from SQL injection, but for most businesses, the defense against SQL-injection attack must be code-based. The opening for SQL-injection attacks comes primarily through Web applications that combine user input with dynamic SQL to form SQL commands that the application sends to the database. Here are four important steps you can take to protect your Web applications from SQL-injection attacks. In addition to the following tips, the Microsoft Patterns and Practices Library that I highlighted last month provides advice about securing your data-access applications.

4. Principle of Least Privilege


The account an application uses to connect to the database should have only the privileges that application requires. The security permissions that an intruder gains from a compromised application define the harm that the intruder can inflict. Applications shouldn't connect as sa or with the Administrator account. Instead, the account should have permissions to access only the database objects it needs.

3. Validate All Input


If an input field should contain numeric data, then verify that users enter only numbers. If character data is acceptable, check for unexpected characters. Make sure your application looks for characters such as semicolons, equals signs, double dashes, brackets, and SQL keywords. The .NET Framework provides regular expressions that enable complex pattern matching, a good way to test user input. Limiting the length of accepted user input is also a good idea. Validating your input might seem obvious, but many applications are vulnerable to SQL-injection attacks because intruders can use the openings that Web applications offer.

2. Avoid Dynamic SQL


Dynamic SQL is a great tool for performing ad hoc queries, but combining dynamic SQL with user input creates exposure that makes SQL-injection attacks possible. Replacing dynamic SQL with prepared SQL or stored procedures is feasible in most applications. Prepared SQL and stored procedures accept user input as parameter data rather than as SQL commands, thus limiting what an intruder can do. Of course, replacing dynamic SQL with a stored procedure won't help you if you use the user input to build dynamic SQL statements in your stored procedures. In that case, the dynamic SQL that the user input creates will still be corrupted, and your database will still be in danger of SQL-injection attack.

1. Use Double Quotes


Replace all the single quotes that your users' input contains with double quotes. This simple precaution will go a long way toward warding off SQL-injection attacks. Single quotes often terminate SQL expressions and give the input more power than is necessary. Replacing the single quotes with double quotes will cause many SQL-injection attacks to fail.