Reviewing Code for Data Validation

One key area in web application security is the validation of data inputted from an external source. Many application exploits are derived from weak input validation on behalf of the application. Weak data validation gives the attacker the opportunity to make the application perform some functionality which it is not meant to do.

How to Avoid Cross-site scripting Vulnerabilities
See the OWASP Development Guide article on Data Validation.

Canonicalization of Input
Input can be encoded to a format that can still be interpreted correctly by the application, but may not be an obvious avenue of attack.

The encoding of ASCII to Unicode is another method of bypassing input validation. Applications rarely test for Unicode exploits and hence provide the attacker a route of attack.

The issue to remember here is that the application is safe if Unicode representation or other malformed representation is input. The application responds correctly and recognises all possible representations of invalid characters.

Example:

The ASCII:

(If we simply block “<” and “>” characters the other representations below shall pass data validation and execute).

URL encoded: %3C%73%63%72%69%70%74%3E

Unicode Encoded: &#60&#115&#99&#114&#105&#112&#116&#62

The OWASP Development Guide delves much more into this subject.

Data Validation Strategy
A general rule is to accept only “Known Good” characters, i.e. the characters that are to be expected. If this cannot be done the next strongest strategy is “Known bad”, where we reject all known bad characters. The issue with this is that today’s known bad list may expand tomorrow as new technologies are added to the enterprise infrastructure.

There are a number of models to think about when designing a data validation strategy, which are listed from the strongest to the weakest as follows.


 * 1) Exact Match (Constrain)
 * 2) Known Good (Accept)
 * 3) Reject Known bad (Reject)
 * 4) Encode Known bad (Sanitise)

In addition, there must be a check for maximum length of any input received from an external source, such as a downstream service/computer or a user at a web browser.

'''Rejected Data must not be persisted to the data store unless it is sanitized. This is a common mistake to log erroneous data, but that may be what the attacker wishes your application to do. '''

e.g.: A Radio button component on a Web page has 3 settings (A, B, C). Only one of those three settings must be accepted (A or B or C). Any other value must be rejected.
 * Exact Match: (preferred method) Only accept values from a finite list of known values.


 * Known Good: If we do not have a finite list of all the possible values that can be entered into the system, we use the known good approach.

e.g.: an email address, we know it shall contain one and only one @. It may also have one or more full stops “.”. The rest of the information can be anything from [a-z] or [A-Z] or [0-9] and some other characters such as “_ “or “–“, so we let these ranges in and define a maximum length for the address.


 * Reject Known bad: We have a list of known bad values we do not wish to be entered into the system. This occurs on freeform text areas and areas where a user may write a note. The weakness of this model is that today known bad may not be sufficient for tomorrow.


 * Encode Known Bad: This is the weakest approach. This approach accepts all input but HTML encodes any characters within a certain character range. HTML encoding is done so if the input needs to be redisplayed the browser shall not interpret the text as script, but the text looks the same as what the user originally typed.

HTML-encoding and URL-encoding user input when writing back to the client. In this case, the assumption is that no input is treated as HTML and all output is written back in a protected form. This is sanitisation in action.

Data Validation Examples
A good example of a pattern for data validation to prevent OS injection in PHP applications would be as follows:

$string = preg_replace("/[^a-zA-Z0-9]/", "", $string);

This code above would replace any non alphanumeric characters with “”. preg_grep could also be used for a True or False result. This would enable us to let “only known good” characters into the application.

Using regular expressions is a common method of restricting input character types. A common mistake in the development of regular expressions is not escaping characters, which are interpreted as control characters, or not validating all avenues of input.

Examples of regular expression are as follows:

http://www.regxlib.com/CheatSheet.aspx

^[a-zA-Z]+$ 	Alpha characters only, a to z and A to Z (RegEx is case sensitive). ^[0-9]+$ 	Numeric only (0 to 9). [abcde] 	Matches any single character specified in set [^abcde] 	Matches any single character not specified in set

Framework Example: (Struts 1.2)
In the J2EE world the struts framework (1.1) contains a utility called the commons validator. This enables us to do two things.


 * 1) Enables us to have a central area for data validation.
 * 2) Provides us with a data validation framework.

What to look for when examining struts is as follows:

The struts-config.xml file must contain the following:

  

This tells the framework to load the validator plug-in. It also loads the property files defined by the comma-separated list. By default a developer would add regular expressions for the defined fields in the validation.xml file.

Next we look at the form beans for the application. In struts, form beans are on the server side and encapsulate the information sent to the application via a HTTP form. We can have concrete form beans (built in code by developers) or dynamic form beans. Here is a concrete bean below:

package com.pcs.necronomicon import org.apache.struts.validator.ValidatorForm; public class LogonForm extends ValidatorForm { private String username; private String password; public String getUsername { return username; }       public void setUsername(String username) { this.username = username; }     public String getPassword { return password; } public void setPassword(String password) { this.password = password; } }

Note the LoginForm extends the ValidatorForm; this is a must as the parent class (ValidatorForm) has a validate method which is called automatically and calls the rules defined in validation.xml.

Now to be assured that this form bean is being called, we look at the struts-config.xml file: It should have something like the following:

  

Next we look at the validation.xml file. It should contain something similar to the following:

    

Note the same name in the validation.xml, the struts-config.xml, this is an important relationship and is case sensitive.

The field “username” is also case sensitive and refers to the String username in the LoginForm class.

The “depends” directive dictates that the parameter is required. If this is blank the error defined in Application.properties. This configuration file contains error messages among other things. It is also a good place to look for information leakage issues.

Error Messages for Validator Framework Validations
errors.required={0} is required. errors.minlength={0} cannot be less than {1} characters. errors.maxlength={0} cannot be greater than {2} characters. errors.invalid={0} is invalid. errors.byte={0} must be a byte. errors.short={0} must be a short. errors.integer={0} must be an integer. errors.long={0} must be a long.0. errors.float={0} must be a float. errors.double={0} must be a double. errors.date={0} is not a date. errors.range={0} is not in the range {1} through {2}. errors.creditcard={0} is not a valid credit card number. errors.email={0} is an invalid e-mail address. prompt.username = User Name is required.

The error defined by arg0, prompt.username is displayed as an alert box by the struts framework to the user. The developer would need to take this a step further by validating the input via regular expression:   mask ^[0-9a-zA-Z]*$</var-value> </form-validation>

Here we have added the Mask directive, this specifies a variable and a regular expression. Any input into the username field which has anything other than A to Z, a to z, or 0 to 9 shall cause an error to be thrown. The most common issue with this type of development is either the developer forgetting to validate all fields or a complete form. The other thing to look for is incorrect regular expressions, so learn those RegEx’s kids!!!

We also need to check if the JSP pages have been linked up to the validation.xml finctionaltiy. This is done by <html:javascript> custom tag being included in the JSP as follows:

<html:javascript formName="logonForm" dynamicJavascript="true" staticJavascript="true" />

Framework Example:(.NET)
The ASP .NET framework contains a validator framework, which has made input validation easier and less error prone than in the past. The validation solution for .NET also has client and server side functionality akin to Struts (J2EE). What is a validator? According to the Microsoft (MSDN) definition it is as follows:

"A validator is a control that checks one input control for a specific type of error condition and displays a description of that problem."

The main point to take out of this from a code review perspective is that one validator does one type of function. If we need to do a number of different checks on our input we need to use more than one validator.

The .NET solution contains a number of controls out of the box: The following is an example web page (.aspx) containing validation:
 * RequiredFieldValidator – Makes the associated input control a required field.
 * CompareValidator – Compares the value entered by the user into an input control with the value entered into another input control or a constant value.
 * RangeValidator – Checks if the value of an input control is within a defined range of values.
 * RegularExpressionValidator – Checks user input against a regular expression.

Validate me baby! <asp:ValidationSummary runat=server HeaderText="There were errors on the page:" /> Please enter your User ID          <asp:RequiredFieldValidator runat=server ControlToValidate=Name ErrorMessage="User ID is required."> * </asp:RequiredFieldValidator> User ID: <input type=text runat=server id=Name> <asp:RegularExpressionValidator runat=server display=dynamic controltovalidate="Name" errormessage="ID must be 6-8 letters." validationexpression="[a-zA-Z0-9]{6,8}" /> <input type=submit runat=server id=SubmitMe value=Submit>

Remember to check that regular expressions are sufficient to protect the application. The “runat” directive means this code is executed at the server prior to being sent to client. When this is displayed to a user’s browser the code is simply HTML.

Example: Classic ASP
There is not built-in validation in classic ASP pages, however you can use regular expressions to accomplish the task. Here is an example of a function with regular expressions to validate a US Zip code.

Public Function IsZipCode (ByVal Text) Dim re  set re = new RegExp re.Pattern = "^\d{5}$" IsZipCode = re.Test(Text) End Function

Length Checking
Another issue to consider is input length validation. If the input is limited by length, this reduces the size of the script that can be injected into the web app.

Many web applications use operating system features and external programs to perform their functions. When a web application passes information from an HTTP request through as part of an external request, it must be carefully data validated for content and min/max length. Without data validation the attacker can inject Meta characters, malicious commands, or command modifiers masquerading as legitimate information, and the web application will blindly pass these on to the external system for execution.

Checking for minimum and maximum length is of paramount importance, even if the code base is not vulnerable to buffer overflow attacks.

If a logging mechanism is employed to log all data used in a particular transaction, we need to ensure that the payload received is not so big that it may affect the logging mechanism. If the log file is sent a very large payload it may crash. Or if it is sent a very large payload repeatedly, the hard disk of the app server may fill, causing a denial of service. This type of attack can be used to recycle the log file, hence removing the audit trail. If string parsing is performed on the payload received by the application, and an extremely large string is sent repeatedly to the application, the CPU cycles used by the application to parse the payload may cause service degradation or even denial of service.

Never Rely on Client-Side Data Validation
Client-side validation can always be bypassed.

Server-side code should perform its own validation. What if an attacker bypasses your client, or shuts off your client-side script routines, for example, by disabling JavaScript?

Use client-side validation to help reduce the number of round trips to the server but do not rely on it for security.

Remember: Data validation must be always done on the server side.

'''A code review focuses on server side code. Any client side security code is not and cannot be considered security.'''

Data validation of parameter names
When data is passed to a method of a web application via HTTP the payload is passed in a “key-value” pair such as  UserId =3o1nk395y  password=letMeIn123

Previously we talked about input validation of the payload (parameter value) being passed to the application. But we also may need to check that the parameter name (UserId, password from above) have not been tampered with.

Invalid parameter names may cause the application to crash or act in an unexpected way. The best approach is “Exact Match” as mentioned previously.

Web Services Data Validation
The recommended input validation technique for web services is to use a schema. A schema is a “map” of all the allowable values that each parameter can take for a given web service method. When a SOAP message is received by the web services handler, the schema pertaining to the method being called is “run over” the message to validate the content of the soap message. There are two types of web service communication methods; XML-IN/XML-OUT and REST (Representational State Transfer). XML-IN/XML-OUT means that the request is in the form of a SOAP message and the reply is also SOAP. REST web services accept a URI request (Non XML) but return a XML reply. REST only supports a point-to-point solution wherein SOAP chain of communication may have multiple nodes prior to the final destination of the request. Validating REST web services input is the same as validating a GET request. Validating an XML request is best done with a schema.

<?xml version="1.0"?> <xsd:schema xmlns:xsd="http://www.w3.org/2001/XMLSchema" xmlns="http://server.test.com" targetNamespace="http://server.test.com" elementFormDefault="qualified" attributeFormDefault="unqualified"> <xsd:complexType name="AddressIn"> <xsd:sequence> <xsd:element name="addressLine1" type="HundredANumeric" nillable="true"/> <xsd:element name="addressLine2" type="HundredANumeric" nillable="true"/> <xsd:element name="county" type="TenANumeric" nillable="false"/> <xsd:element name="town" type="TenANumeric" nillable="true"/> <xsd:element name="userId" type="TenANumeric" nillable="false"/> </xsd:sequence> </xsd:complexType> <xsd:simpleType name="HundredANumeric"> <xsd:restriction base="xsd:string"> <xsd:minLength value="1"/> <xsd:maxLength value="100"/> <xsd:pattern value="[a-zA-Z0-9]"/> </xsd:restriction> </xsd:simpleType> <xsd:simpleType name="TenANumeric"> <xsd:restriction base="xsd:string"> <xsd:minLength value="1"/> <xsd:maxLength value="10"/> <xsd:pattern value="[a-zA-Z0-9]"/> </xsd:restriction> </xsd:simpleType> </xsd:schema>

Here we have a schema for an object called AddressIn. Each of the elements has restrictions applied to it and the restrictions (in red) define what valid characters can be inputted into each of the elements. What we need to look for is that each of the elements has a restriction applied to it, as opposed to the simple type definition such as xsd:string. This schema also has the <xsd:sequence> tag applied to enforce the sequence of the data that is to be received.

Example-Perl
The following snippet of Perl code demonstrates code which is vulnerable to XSS.

use CGI; my $cgi = CGI->new; my $value = $cgi->param('value'); print $cgi->header; print "You entered $value";
 * 1) !/usr/bin/perl

The code blindly accepts data supplied in the parameter labeled 'value'. To add to this problem of accepting data with no validation, the code will display the inputted data to the user. If you have read this far into the paper I hope the light bulb is now flashing above your head with the realisation that this particular vulnerability would allow a Reflected XSS attack to occur.

The 'value' parameter should validate the supplied data and only print data which has been 'cleaned' by the validation filter. There are multiple options available with Perl to validate this parameter correctly. Firstly, a simple and crude filter is shown below:

$value =~ s/[^A-Za-z0-9 ]*/ /g;

This will restrict the data in the parameter to uppercase, lowercase, spaces, and numbers only. This of course removes the dangerous characters we have associated with XSS such as.

A second option would be to use the HTML::Entities module for Perl which will force HTML encoding on the inputted data. I have changed the code to incorporate the HTML::Entities module and given an example out the encoding in action.

use CGI; use HTML::Entities; my $cgi = CGI->new; my $value = $cgi->param('value'); print $cgi->header; print "You entered ", HTML::Entities::encode($value);
 * 1) !/usr/bin/perl

If the data provided was <SCRIPT>alert(“XSS”)</SCRIPT> the HTML::Entities module would produce the following output:

&amp;lt;SCRIPT&amp;gt;alert(&amp;quot;XSS&amp;quot;)&amp;lt;/SCRIPT&amp;gt;

This would remove the threat posed by the original input.

Example-PHP
PHP allows users to create dynamic web pages quite easily, and this led to many implementations of PHP which lacked any security thought.

The example provided below shows very simple PHP message board which has been setup without sufficient data validation.

<input type="text" name="inputs">

<?php if (isset($_GET['inputs'])) {    $fp = fopen('./inputs.txt', 'a'); fwrite($fp, "{$_GET['inputs']} "); fclose($fp); } readfile('./inputs.txt'); ?>

You can see that this simple form takes the user inputs and writes it to the file named inputs.txt.

This file is then used to write the message to the message board for other users to see. The danger posed by this form should be clear straight away, the initial input is not subject to any kind of validation and is presented to other users as malicious code.

This could have been avoided by implementing simple validation techniques. PHP allows the developer to use the htmlentities function. I have added the htmlentities to the form:

<input type="text" name="inputs">

<?php if (isset($_GET['inputs'])) {    $message = htmlentities($_GET['inputs']); $fp = fopen('./inputs.txt', 'a'); fwrite($fp, "$inputs "); fclose($fp); } readfile('./inputs.txt'); ?>

The addition is simple but the benefits gained can be substantial. The messageboard now has some protection against any script code that could have been entered by a malicious user. The code will now be HTML entity encoded by the htmlentities function.

Example–Classic ASP
Just like in PHP, ASP pages allow dynamic content creation, so for an XSS vulnerable code like the following:

Response.Write "Please confirm your name is " & Request.Form("UserFullName") We will use the HTMLEncode Built-in function in the following way: Response.Write "Please confirm your name is " & Server.HTMLEncode (Request.Form("UserFullName"))

Example–JavaScript
The fourth and final example we will look at is JavaScript code. Again we will show a vulnerable piece of code and then the same code with data validation in place.

We will observe some vulnerable JavaScript which takes the user’s name from the URL and uses this to create a welcome message. The vulnerable script is displayed below:

<SCRIPT> var pos=document.URL.indexOf("name=")+5; document.write(document.URL.substring(pos,document.URL.length)); </SCRIPT>

The problem with this script was discussed earlier; there is no validation of the value provide for “name=”.

I have fixed the script below using a very simple validation technique.

<SCRIPT> var pos=document.URL.indexOf("name=")+5; var name=document.URL.substring(pos,document.URL.length); if (name.match(/^[a-zA-Z]$/)) {  document.write(name); } else { window.alert("Invalid input!"); } </SCRIPT>

The 3rd line of the script ensures that the characters are restricted to uppercase and lowercase for the user name. Should the value provided violate this, an invalid input error will be returned to the user.