Category:OWASP JSP Testing Tool Project
Summer of Code 2008 Status
|Project Name||OWASP JSP Testing Tool Project|
|Short Project Description||The goal of this project is to create an easy to use, freely available tool that can be used to quickly ascertain the level of protection that each component of a JSP tag library offers. This information can serve two purposes: |
|Short Project Description||Jason Li||Project Contributors
New BSD License
OWASP SoC 08
|Release Status||Main Links||Related Projects|
If any, add link here
About the Project
Cross-site scripting is one of the most pervasive web application security vulnerabilities present in today's applications, ranking at the top of the most recent OWASP Top Ten (2007). Recently, UI frameworks have emerged that allow application developers to create web components quickly and effortlessly. Java EE facilitates this ability by providing JSP tags that can be conglomerated as a tag library, such as was done with Java Server Faces (JSF). These tags allow developers to create rich user interfaces by parameterizing attributes of the web component. With the advent of this design, a unique opportunity arises to prevent cross-site scripting. If tag libraries offered built-in protection from cross-site scripting, then using parameterized web components to prevent cross-site scripting could become analogous to using parameterized queries to prevent SQL injection. However, security is not always in mind when such tag libraries are developed and thus it is not clear which libraries, if any, offer such built-in protection. This project would allow developers to quickly determine the protection mechanisms provided by any tag libraries they develop or of any third-party tag libraries they are considering using for their application. This will enable a push to parameterize web components which will serve to make web applications safer.
This project was born out of an idea while supporting a client. The client had created their own custom UI components based on the core JSF components. Developers were encouraged to use this custom library as it was thought to be “safe” from cross-site scripting. After the library gained popularity internally, the client decided they should verify the library did indeed protect web developers from cross-site scripting. This testing required a lot of manual generation of test pages in order to instantiate the various tags and thoroughly exercise each of the various attributes. We realized that the concept of having a safe UI component library is a sound idea and potentially popular across many organizations. Hence, the idea was born to create a tool that could test custom UI tag libraries to measure the level of protection from cross-site scripting offered by the library. This project is the implementation of that idea.
About the Author
Jason is an Application Security Engineer at Aspect Security during which time he has performed code reviews, penetration testing and training at a variety of financial, commercial, and government institutions. He is a certified GIAC Secure Software Programmer in Java and before joining Aspect, he was a Java Software Developer and a Java course instructor for Johns Hopkins University. He is a core developer on the OWASP AntiSamy Project. Jason eared his Post-Master's in Computer Science with a concentration in Informaiton Security from Johns Hopkins University. He previously earned his his Master's and B.S from Cornell University (both in Computer Science).
This project uses a report in order to show the level of protection offered by a library. The goal of the report is for developers to quickly ascertain what protections a custom tag library offers. In addition, the report can be used to identify specific tags or attributes that are not safe and need a different protection mechanism (for example, enforcing a particular code pattern or practice).
This information is displayed in tabular format with each individual tag having its own table. Each row is an attribute of the tag, except for the first row which is reserved for testing the text content of a tag (for example, content). Each column is a test attack that is attempted within the tag. Each corresponding cell’s color indicates the status of that particular attribute against that test attack with red indicating the test failed, green indicating the test passed and yellow indicating an error occurred during testing.
How the Report Works
testcase.vm) to provide isolation. Each test attack attempts to execute a script function named
x() that is defined on the test page. The function populates an element on the page with an indicator that the test either passed or failed. The overall report page is a generated HTML page that contains the table described above and hidden IFRAME with every test page. Once the pages load, the report page evaluates the status indicated in the frame and changes the color of the corresponding table cell accordingly. Note that clicking on any table cell will toggle the visibility of the corresponding test case's IFRAME.
The web application configures an error page (
error.jsp) which handles the inevitable application errors from improper usage of custom tags. The error page generates a status indicator that includes the exception stack trace, both of which can be read by the containing report page. In addition to appearing yellow, the title attribute of the corresponding table cell is populated with the exception stack trace and will appear in most browsers as a tooltip. As the project matures, some mechanism will be provided to address tag generation errors (see later section).
Note that because this report strategy relies heavily on a web browser, the results may differ from browser to browser. For example, the CSS based attack renders successfully in Internet Explorer but not in Firefox. The report could be made browser agnostic by instead searching through the generated JSP content. That approach has tradeoffs as it raises the possibility of false positives. By keeping the browser involved, organizations can test against browsers they support and be assured that if the attack test case executed, it is an issue.
How the Report is Generated
The report generation is broken down into three phases:
- Parsing of the Tag Library Document (TLD)
- Generation of Test Pages
- Deploying of Report
Parsing of the Tag Library Document
Ideally, this project would have leveraged existing code to parse TLDs. Application servers routinely parse TLDs in order to compile JSPs that use tag libraries and several application servers are open source (Tomcat, JBoss, etc). However, upon further investigation, most existing implementations were tightly intertwined with the rest of the application server. Nonetheless, the core Java EE API includes several classes to encapsulate tag information (see the javax.servlet.jsp.tagext package). This project used as many of these existing classes as possible to both mitigate the volume of work and also to permit leveraging existing application server code in the future. The code uses JAXP, which has been implemented in standard Java since version 1.4, to parse TLD documents.
Generation of Test Pages
The code uses a report and test case template that is populated using Apache Velocity. Using the encapsulated TLD metadata from the parsed TLD, the code iterates through all tags and attributes to construct test case pages for each test attack. The test attacks come from an XML file (
attacks.xml) which represents a repository of various cross-site scripting attacks that invoke the
Deploying the Report
After the report HTML and test case JSPs are generated, they must be deployed to a JSP container (typically an application server). The container should be configured with any mappings and JAR libraries required to use the custom tag libraries.
These ideas are features that are recognized as desirable but will not be considered until the tool can test a reasonable JSP tag library. Currently, that measuring stick is the JSF Core library.
Programmatic Tag Generation
One of the limitations in the current tag generation is that tags are generated relatively blindly. This strategy can result in errors when supplied tag attributes do not conform to required formats. One possible strategy is to integrate a JSP/Servlet container where tags can be generated and rendered programmatically. This approach was deemed too complicated for the first iteration of this tool.
Another issue is that different permutations of attributes may result in different behavior which can affect the outcome of a test case. This issue can be resolved by generating multiple permutations of attributes, but this becomes infeasible as the number of attributes grows.
Pages in category "OWASP JSP Testing Tool Project"
This category contains only the following page.