A1.1 Description

Web applications use input from HTTP requests (and occasionally files) to determine how to respond. Attackers can tamper with any part of an HTTP request, including the url, querystring, headers, cookies, form fields, and hidden fields, to try to bypass the site's security mechanisms. Common names for common input tampering attacks include: forced browsing, command insertion, cross site scripting, buffer overflows, format string attacks, SQL injection, cookie poisoning, and hidden field manipulation. Each of these attack types is described in more detail later in this paper.

• A4: Cross Site Scripting Flaws discusses input that contains scripts to be executed on other user's browsers
• A5: Buffer Overflows discusses input that has been designed to overwrite program execution space
• A6: Injection Flaws discusses input that is modified to contain executable commands

Some sites attempt to protect themselves by filtering out malicious input. The problem is that there are so many different ways of encoding information. Since almost all HTTP input can be represented in multiple formats, unusual encodings can be used to obfuscate any attack targeting the vulnerabilities described in this document. This makes filtering very difficult. To protect themselves, web applications should convert input parameters to the simplest form before they are validated, otherwise, malicious input can be masked and slip past filters. The process of simplifying these encodings is called "canonicalization."

A surprising number of web applications use only client-side mechanisms to validate input. Client side validation mechanisms are easily bypassed, leaving the web application without any protection against malicious parameters. Attackers can generate their own HTTP requests using tools as simple as telnet. They do not have to pay attention to anything that the developer intended to happen on the client side. Client side validation is a fine idea for performance and usability, but it has no security benefit whatsoever. Server side checks are required to defend against parameter manipulation attacks. Once these are in place, client side checking can also be included to enhance the user experience for legitimate users and/or reduce the amount of invalid traffic to the server.

These attacks are becoming increasingly likely as the number of tools that support parameter "fuzzing", corruption, and brute forcing grows. The impact of using unvalidated input should not be underestimated. A huge number of attacks would become difficult or impossible if developers would simply validate input before using it. Unless a web application has a strong, centralized mechanism for validating all input from HTTP requests (and any other sources), vulnerabilities based on malicious input are very likely to exist.

A1.2 Environments Affected

All web servers, application servers, and web application environments are susceptible to parameter tampering.

The following table lists some of the available tools for checking for unvalidated input across different development platforms.

A1.3 Examples and References

• OWASP Guide to Building Secure Web Applications and Web Services: Data Validation

• modsecurity project (Apache module for HTTP validation)

• How to Build an HTTP Request Validation Engine (J2EE validation with Stinger)

• Have Your Cake and Eat it Too (.NET validation)

A1.4 How to Determine If You Are Vulnerable

Any part of an HTTP request that is used by a web application without being carefully validated is known as a "tainted" parameter. The simplest way to find tainted parameter use is to have a detailed code review, searching for all the calls where information is extracted from an HTTP request. For example, in a J2EE application, these are the methods in the HttpServletRequest class. Then you can follow the code to see where that variable gets used. If the variable is not checked before it is used, there is very likely a problem. In Perl, you should consider using the "taint" (-T) option.

It is also possible to find tainted parameter use by using tools like OWASP's WebScarab. By submitting unexpected values in HTTP requests and viewing the web application's responses, you can identify places where tainted parameters are used.

Specific examples include:

• Global Variables Exploit in PHP with Post and Get - If PHP is configured with "set globals on" (the default setting previous to version 4.3), application variables can be initialized via an HTTP request. An application may be vunerable to unexpected values passed by the user.

• Cross-site Scripting - See section A4

• SQL Injection - use variable binding if possible when executing queries. Otherwise, escape all strings and treat integers as strings by enclosing them in quotes.

• URL manipulation - Validate URLs that are entered in forms.

• Embedded commands in echoed data - Encode HTML special characters before echoing them back to the browser.

A1.5 How to Protect Yourself

The best way to prevent parameter tampering is to ensure that all parameters are validated before they are used. A centralized component or library is likely to be the most effective, as the code performing the checking should all be in one place. Each parameter should be checked against a strict format that specifies exactly what input will be allowed. "Negative" approaches that involve filtering out certain bad input or approaches that rely on signatures are not likely to be effective and may be difficult to maintain.

Parameters should be validated against a "positive" specification that defines:

• Data type (string, integer, real, etc...)

• Allowed character set

• Minimum and maximum length

• Whether null is allowed

• Whether the parameter is required or not

• Whether duplicates are allowed

• Numeric range

• Specific legal values (enumeration)

• Specific patterns (regular expressions)

A new class of security devices known as web application firewalls can provide some parameter validation services. However, in order for them to be effective, the device must be configured with a strict definition of what is valid for each parameter for your site. This includes properly protecting all types of input from the HTTP request, including URLs, forms, cookies, querystrings, hidden fields, and parameters.

The OWASP Filters project is producing reusable components in several languages to help prevent many forms of parameter tampering. The Stinger HTTP request validation engine (http://www.aspectsecurity.com/stinger/ ) was also developed by OWASP for J2EE environments.