OWASP Benchmark Project

• Main
• Test Cases
• Tool Results
• Documentation
• Acknowledgements

OWASP Benchmark Project The OWASP Benchmark for Security Automation (OWASP Benchmark) is a test suite designed to evaluate the speed, coverage, and accuracy of automated vulnerability detection tools and services (henceforth simply referred to as 'tools'). Without the ability to measure these tools, it is difficult to understand their value or interpret vendor claims. The OWASP Benchmark contains over 20,000 test cases that are fully runnable and exploitable. You can use the OWASP Benchmark with Static Application Security Testing (SAST) tools. A future goal is to support the evaluation of Dynamic Application Security Testing (DAST) tools like OWASP ZAP and Interactive Application Security Testing (IAST) tools. The current version of the Benchmark is implemented in Java. Future versions may expand to include other languages. Benchmark Project Philosophy Security tools (SAST, DAST, and IAST) are amazing when they find a complex vulnerability in your code. But they can drive everyone crazy with complexity, false alarms, and missed vulnerabilities. Using these tools without understanding their strengths and weaknesses can lead to a dangerous false sense of security. We are on a quest to measure just how good these tools are at discovering and properly diagnosing security problems in applications. We rely on the long history of military and medical evaluation of detection technology as a foundation for our research. Therefore, the test suite tests both real and fake vulnerabilities. There are four possible test outcomes in the Benchmark: Tool correctly identifies a real vulnerability (True Positive - TP) Tool fails to identify a real vulnerability (False Negative - FN) Tool correctly ignores a false alarm (True Negative - TN) Tool fails to ignore a false alarm (False Positive - FP) We can learn a lot about a tool from these four metrics. A tool that simply flags every line of code as vulnerable will perfectly identify all vulnerabilities in an application, but will also have 100% false positives. Similarly, a tool that reports nothing will have zero false positives, but will also identify zero real vulnerabilities. Imagine a tool that flips a coin to decide whether to report each vulnerability for every test case. The result would be 50% true positives and 50% false positives. We need a way to distinguish valuable security tools from these trivial ones. If you imagine the line that connects all these points, from 0,0 to 100,100 establishes a line that roughly translates to "random guessing." The ultimate measure of a security tool is how much better it can do than this line. The diagram below shows how we will evaluate security tools against the Benchmark. Benchmark Validity The Benchmark tests are not exactly like real applications. The tests are derived from coding patterns observed in real applications, but many of them are considerably simpler than real applications. Other tests may have coding patterns that don't occur frequently in real code. It's best to imagine the Benchmark as a continuum of tests from very simple all the way up to pretty difficult. Remember, we are trying to test the capabilities of the tools and make them explicit, so that *users* can make informed decisions about what tools to use, how to use them, and what results to expect. This is exactly aligned with the OWASP mission to make application security visible. Benchmark Scoring and Reporting Results We encourage both vendors, open source tools, and end users to verify their application security tools against the Benchmark. We encourage everyone to contribute their results to the project. In order to ensure that the results are fair and useful, we ask that you follow a few simple rules when publishing results. We won't recognize any results that aren't easily reproducible. Provide an easily reproducible procedure (script preferred) to run the tool on the Benchmark, including: A description of the default “out-of-the-box” installation, version numbers, etc… All configuration, tailoring, onboarding, etc… performed to make the tool run All changes to default security rules, tests, or checks used to achieve the results Easily reproducible steps to run the tool Scripts to generate results in the format below Results should be in the following table format and provide details for each category of tests, as well as overall statistics. Security Category TP FN TN FP Total TPR FPR Score General security category for test cases True Positives: Tests with real vulnerabilities that were correctly reported as vulnerable by the tool False Negative: Tests with real vulnerabilities that were not correctly reported as vulnerable by the tool True Negative: Tests with fake vulnerabilities that were correctly not reported as vulnerable by the tool False Positive:Tests with fake vulnerabilities that were incorrectly reported as vulnerable by the tool Total test cases for this category True Positive Rate: TP / ( TP + FN ) False Positive Rate: FP / ( FP + TN ) Normalized distance from the “guess line” TPR - FPR Command Injection ... ... ... ... ... ... ... ... Etc... ... ... ... ... ... ... ... ... Total TP Total FN Total TN Total FP Total TC Average TPR Average FPR Average Score Code Repo The code for this project is hosted at the OWASP Git repository. Along with the code comes a Maven pom.xml file so you can download all the dependencies and build the entire project with ease using Maven. Using the pom, it should be easy to verify all the code compiles correctly. To download and build everything, if you already have git and maven installed, all you have to do is: $ git clone https://github.com/OWASP/benchmark $ cd benchmark $ mvn compile Licensing The OWASP Benchmark is free to use under the GNU General Public License v2.0. Mailing List OWASP Benchmark Mailing List Project Leaders Dave Wichers @ Project References Software Assurance Marketplace (SWAMP) - set of curated packages to test tools against SAMATE List of Test Collections Related Projects NSA's Juliet for Java WAVESEP

Quick Download All test code and project files can be downloaded from OWASP GitHub. News and Events April 15, 2015 - Benchmark Version 1.0 Released May 23, 2015 - Benchmark Version 1.1 Released Classifications GNU General Public License v2.0

Version 1.0 of the Benchmark was published on April 15, 2015 and had 20,983 test cases. On May 23, 2015, version 1.1 of the Benchmark was released. The 1.1 release improves on the previous version by making sure that there are both true positives and false positives in every vulnerability area. The test case areas and quantities for the 1.1 release are:

Vulnerability Area	Number of Tests	CWE Number
Command Injection	2708	78
Weak Cryptography	1440	327
Weak Hashing	1421	328
LDAP Injection	736	90
Path Traversal	2630	22
Secure Cookie Flag	416	614
SQL Injection	3529	89
Trust Boundary Violation	725	501
Weak Randomness	3640	330
XPATH Injection	347	643
XSS (Cross-Site Scripting)	3449	79
Total Test Cases	21,041

To download a spreadsheet that lists every test case, the vulnerability category, the CWE number, and the expected result (true finding/false positive), click here.

Every test case is:

• a servlet or JSP (currently they are all servlets, but we plan to add JSPs soon)
• either a true vulnerability or a false positive for a single issue

The benchmark is intended to help determine how well analysis tools correctly analyze a broad array of application and framework behavior, including:

• HTTP request and response problems
• Simple and complex data flow
• Simple and complex control flow
• Popular frameworks
• Inversion of control
• Reflection
• Class loading
• Annotations
• Popular UI technologies (particularly JavaScript frameworks)

Not all of these are yet tested by the Benchmark but future enhancements intend to provide more coverage of these issues.

Additional future enhancements could cover:

• All vulnerability types in the OWASP Top 10
• Does the tool find flaws in libraries?
• Does the tool find flaws spanning custom code and libraries?
• Does tool handle web services? REST, XML, GWT, etc…
• Does tool work with different app servers? Java platforms?

Example Test Case

Each test case is a simple Java EE servlet. BenchmarkTest00001 in version 1.0 of the Benchmark was an LDAP Injection test with the following metadata in the accompanying BenchmarkTest00001.xml file:

 <test-metadata>
   <category>ldapi</category>
   <test-number>00001</test-number>
   <vulnerability>true</vulnerability>
   <cwe>90</cwe>
 </test-metadata>

BenchmarkTest00001.java in the OWASP Benchmark 1.0 simply reads in all the cookie values, looks for a cookie named "foo", and uses the value of this cookie when performing an LDAP query. Here's the code for BenchmarkTest00001.java:

 package org.owasp.benchmark.testcode;
 
 import java.io.IOException;
 
 import javax.servlet.ServletException;
 import javax.servlet.annotation.WebServlet;
 import javax.servlet.http.HttpServlet;
 import javax.servlet.http.HttpServletRequest;
 import javax.servlet.http.HttpServletResponse;
 
 @WebServlet("/BenchmarkTest00001")
 public class BenchmarkTest00001 extends HttpServlet {
 	
 	private static final long serialVersionUID = 1L;
 	
 	@Override
 	public void doGet(HttpServletRequest request, HttpServletResponse response) throws ServletException, IOException {
 		doPost(request, response);
 	}
 
 	@Override
 	public void doPost(HttpServletRequest request, HttpServletResponse response) throws ServletException, IOException {
 		// some code
 
 		javax.servlet.http.Cookie[] cookies = request.getCookies();
 		
 		String param = null;
 		boolean foundit = false;
 		if (cookies != null) {
 			for (javax.servlet.http.Cookie cookie : cookies) {
 				if (cookie.getName().equals("foo")) {
 					param = cookie.getValue();
 					foundit = true;
 				}
 			}
 			if (!foundit) {
 				// no cookie found in collection
 				param = "";
 			}
 		} else {
 			// no cookies
 			param = "";
 		}
 		
 		try {
 			javax.naming.directory.DirContext dc = org.owasp.benchmark.helpers.Utils.getDirContext();
 			Object[] filterArgs = {"a","b"};
 			dc.search("name", param, filterArgs, new javax.naming.directory.SearchControls());
 		} catch (javax.naming.NamingException e) {
 			throw new ServletException(e);
 		}
 	}
 }

The results for 3 free tools, PMD, FIndBugs, and FindBugs with the FindSecBugs plugin are now available here: https://cdn.rawgit.com/OWASP/Benchmark/master/scorecard/OWASP_Benchmark_Home.html

We have computed results for all the following tools, but are prohibited from publicly releasing the results for any commercial tools due to their licensing restrictions against such publication.

The Benchmark can generate results for the following tools:

Free Static Analysis Tools:

• PMD (which really has no security rules)
• Findbugs
• FindBugs with the FindSecurityBugs plugin
• SonarQube

Commercial Static Analysis Tools:

• Coverity Code Advisor (On-Demand and stand-alone versions)
• HP Fortify (On-Demand and stand-alone versions)
• IBM AppScan Source
• Parasoft Jtest
• Veracode SAST

The free tools come bundled with the Benchmark so you can run them yourselves. If you have a license for any commercial SAST tool, you can also run them against the Benchmark. Just put your results files in the /results folder of the project, and then run the BenchmarkScore script for your platform (.sh / .bat) and it will generate a scorecard in the /scorecard directory for all the tools you have results for that are currently supported.

WARNING: If you generate results for a commercial tool, be careful who you distribute it to. Each tool has its own license defining when any results it produces can be released/made public. It is likely to be against the terms of a commercial tool's license to publicly release that tool's score against the OWASP Benchmark. The OWASP Benchmark project takes no responsibility if someone else releases such results. It is for just this reason that the Benchmark project isn't releasing such results itself.

If you would like to contribute to this project by running a tool against the Benchmark and sending us the tool's results file, please contact the project lead.

The project has automated test harnesses for these vulnerability detection tools, so we can repeatably run the tools against each version of the Benchmark and automatically produce scorecards in our desired format.

We want to test as many tools as possible against the Benchmark. If you are:

• A tool vendor and want to participate in the project
• Someone who wants to help score a free tool agains the project
• Someone who has a license to a commercial tool and the terms of the license allow you to publish tool results, and you want to participate

please let me know!

What is in the Benchmark?

The Benchmark is a Java Maven project. Its primary component is thousands of test cases (e.g., BenchmarkTest00001.java) , each of which is a single Java servlet that contains a single vulnerability (either a true positive or false positive). The vulnerabilities span about a dozen different types currently and are expected to expand significantly in the future.

An expectedresults.csv is published with each version of the Benchmark (e.g., expectedresults-1.1.csv) and it specifically lists the expected results for each test case. Here’s what the first two rows in this file looks like for version 1.1 of the Benchmark:

# test name		category	real vulnerability	CWE	Benchmark version: 1.1	2015-05-22
BenchmarkTest00001	crypto		TRUE			327

This simply means that the first test case is a crypto test case (use of weak cryptographic algorithms), this is a real vulnerability (as opposed to a false positive), and this issue maps to CWE 327. It also indicates this expected results file is for Benchmark version 1.1 (produced May 22, 2015). There is a row in this file for each of the tens of thousands of test cases in the Benchmark. Each time a new version of the Benchmark is published, a new corresponding results file is generated and each test case can be completely different from one version to the next.

The Benchmark also comes with a bunch of different utilities, commands, and prepackaged open source security analysis tools, all of which can be executed through Maven goals, including:

• Open source vulnerability detection tools to be run against the Benchmark
• A scorecard generator, which computes a scorecard for each of the tools you have results files for.

What Can You Do With the Benchmark?

• Compile all the software in the Benchmark project (e.g., mvn compile)
• Run a vulnerability detection tool against the Benchmark test cases to generate a results file for that tool.
  • These prepackaged open source tools have Maven goals for running these tools on the Benchmark. The following open source tools are currently supported:
    • PMD
    • FindBugs
    • FindBugs with the FindSecurityBugs extension for FindBugs
    • SonaryQube (support coming soon)

• Generate scorecards for each of the tools you have results files for. In addition to supporting the open source tools listed above, scorecards for the following commercial tools can also be generated:
  • Coverity Code Advisor
  • HP Fortify
  • IBM AppScan Source
  • Parasoft Jtest
  • Veracode SAST

The Benchmark project team can quickly generate new scorecard generators for other static analysis tools. We just don’t have access to any other licenses at this time in order to run them against the Benchmark. If you have a license, and can send us a tool generated results file, please send it to us, and we’ll generate a scorecard generator for it so that tool is now formally supported by the project.

Getting Started

Before downloading or using the Benchmark make sure you have the following installed and configured properly:

GIT: http://git-scm.com/ or https://github.com/
Maven: https://maven.apache.org/ 
Java: http://www.oracle.com/technetwork/java/javase/downloads/index.html (Java 7 or 8)

Getting and building the Benchmark

To download and build everything:

$ git clone https://github.com/OWASP/benchmark 
$ cd benchmark
$ mvn compile

Running Analysis Tools Against the Benchmark

There are scripts for running each of the free vulnerability detection tools included with the Benchmark against the Benchmark test cases. On Linux, you might have to make them executable (e.g., chmod 755 scriptname) before you can run them.

Generating Test Results for PMD:

$ runPMD.sh (Linux) or runPMD.bat (Windows)

Generating Test Results for FindBugs:

$ runFindBugs.sh (Linux) or runFindBugs.bat (Windows)

Generating Test Results for FindBugs with the FindSecBugs plugin:

$ runFindSecBugs.sh (Linux) or runFindSecBugs.bat (Windows)

In each case, the script will generate a results file and put it in the /results directory. For example:

Benchmark_1.1-findbugs-v3.0.1-1026.xml

This results file name is carefully constructed to mean the following: It's a results file against the OWASP Benchmark version 1.1, FindBugs was the analysis tool, it was version 3.0.1 of FindBugs, and it took 1026 seconds to run the analysis.

NOTE: If you create a results file yourself, by running a commercial tool for example, you can add the version # and the compute time to the filename just like this and the Benchmark Scorecard generator will pick this information up and include it in the generated scorecard. If you don't, depending on what metadata is included in the tool results, the Scorecard generator might do this automatically anyway.

Generating Scorecards

The application BenchmarkScore is included with the Benchmark. It parses the output files generated by security tools run against the Benchmark and compares them against the expected results, and produces a set of web pages that detail the accuracy and speed of the tools involved.

BenchmarkScore includes parsers for results files from these tools:

• FindBugs (.xml)
• FindSecBugs (.xml)
• PMD (.xml)
• SonarQube (.xml)
• Coverity Code Advisor (.json)
• HP Fortify (both .fpr and .fvdl files)
• IBM Appscan (.ozasmt)
• Parasoft Jtest (.xml)
• Veracode SAST (.xml)

The following command will compute a BenchmarkScore for all the results files in the /results directory:

createScorecard.sh (Linux) or createScorecard.bat (Windows)

We recommend including the BenchmarkScore version number in any results file, in order to help prevent mismatches between the expected results and the actual results files. A tool will not score well against the wrong expected results.

WARNING: If you generate results for a commercial tool, be careful who you distribute it to. Each tool has its own license defining when any results it produces can be released/made public. It is likely to be against the terms of a commercial tool's license to publicly release that tool's score against the OWASP Benchmark. The OWASP Benchmark project takes no responsibility if someone else releases such results. It is for just this reason that the Benchmark project isn't releasing such results itself.

The following people have contributed to this project and their contributions are much appreciated!

• Juan Gama - Development of initial release and continued support
• Ken Prole - Assistance with automated scorecard development using CodeDx
• Nick Sanidas - Development of initial release

We are looking for volunteers. Please contact Dave Wichers if you are interested in contributing new test cases, tool results run against the benchmark, or anything else.