Phosphor is a system for performing dynamic taint tracking in the Java Virtual Machine (JVM), on commodity JVMs (e.g. Oracle's HotSpot or OpenJDK's IcedTea). Phosphor uses Java bytecode instrumentation to associate labels, which we referred to as taint tags, with program data and propagate these labels along information "flows" at runtime. This repository contains the source code for Phosphor. For more information about how Phosphor works and its uses, please refer to our OOPSLA 2014 paper, ISSTA 2015 Tool Demo, or email Jonathan Bell. José Cambronero also maintains a series of examples on using Phosphor.
Phosphor has been extensively developed since its original publication, and now includes many features and options not described in the OOPSLA 2014 paper. If you are looking to replicate our OOPSLA 2014 experiments, the easiest way to get the same version with certainty is to use this VM Image with all relevant files here, and to follow the README with instructions for doing so here.
Phosphor currently requires Java 9+ to build, but it can also be used on Java 8.
This branch contains what is nearly a complete rewrite of Phosphor, using a less fragile (but slower) approach to pass taint tags between methods. It also is the first version of Phosphor to support Java 9+.
Remaining tasks for this branch before promotion:
MultiTainter.taintedObjectexport JAVA_HOME=<PATH-TO-JDK>, where <PATH-TO-JDK> is the path
of the JDK installation.mvn -DskipTests install.Once you have built Phosphor according to the directions described above in the
section "Building Phosphor", you can run Phosphor's tests and examples.
Although Phosphor currently requires Java 9+ to build, it can also be used on Java 8.
If you would like to run Phosphor's tests on Java 8, build Phosphor using Java 9+, then change the
JAVA_HOME environmental variable to the path of a JDK 8 installation before running the tests.
To run Phosphor the root directory of this project, run mvn -pl :integration-tests verify.
The first time you run this command, Maven will invoke the Phosphor Maven plugin to create
Phosphor-instrumented Java installations.
These instrumented Java installation are cached for future use and will not be recreated unless one of the
Phosphor JARs, the configuration used to create them, or the value of JAVA_HOME changes.
Once the Phosphor Maven plugin finishes creating the instrumented Java installations, the tests will run.
These tests demonstrate how Phosphor can be used and are a good reference when first learning Phosphor.
To track the flow of information through classes in the Java Class Library (JCL), such as java.lang.String
and java.util.List, Phosphor must instrument the bytecode of JCL classes.
Therefore, the first step when using Phosphor is to create an instrumented Java installation
(i.e., Java Development Kit or Java Runtime Environment).
A Java installation can be downloaded from Oracle or
the Adoptium Working Group.
Once you have obtained a Java installation, it can be instrumented either using Phosphor's driver or
Maven plugin.
We discuss both options below.
Important note on OpenJDK vs Oracle's Java installations:
Oracle's Java installations require that the JAR that contains the cryptography routines jce.jar be signed by
Oracle for export control purposes.
OpenJDK does not.
Phosphor instrumentation will break these signatures.
Therefore, it is not possible to use Phosphor with Oracle's Java installation and use the cryptography functionality.
The Phosphor driver can be used to apply Phosphor instrumentation to Java classes in a Java installation,
directory, or archive.
If you have built Phosphor according to the directions described above in the
section "Building Phosphor", then the driver JAR will be available at
phosphor-driver/target/phosphor-driver-VERSION.jar relative to the root of this project.
The latest snapshot of the driver JAR is available at the
Sonatype OSS Repository Hosting (OSSRH).
The driver JAR can also be acquired using the Maven dependency:
<build>
...
<dependencies>
<dependency>
<groupId>edu.gmu.swe.phosphor</groupId>
<artifactId>phosphor-driver</artifactId>
<version>VERSION</version>
</dependency>
</dependencies>
...
</build>The driver supports archives of types JAR, WAR, ZIP, and JMOD. The driver takes a list of options followed by a source location and a destination location:
java -jar <phosphor-driver.jar> [OPTIONS] <SOURCE> <DEST>The driver will create a Phosphor-instrumented copy of Java installation, directory, or archive located at the specified source location at place it at the specified destination location. The options you specify allow you to control how Phosphor instruments the specified source. A detailed list of available options can be accessed by running:
java -jar <phosphor-driver.jar> -helpTo create a Phosphor-instrumented Java installation as part of your Maven build, add the
phosphor-instrument-maven-plugin in your pom:
<build>
...
<plugins>
...
<plugin>
<groupId>edu.gmu.swe.phosphor</groupId>
<artifactId>phosphor-instrument-maven-plugin</artifactId>
<version>VERSION</version>
</plugin>
...
</plugins>
...
</build>See the documentation for the InstrumentingMojo for more detailing on how to configure and use the Phosphor Maven plugin.
If you choose, you can also use the Phosphor driver to instrument you application classes before running your
application.
This step is optional;
Phosphor will dynamically instrument any classes not already instrumented at runtime as they are loaded by the JVM.
If you plan to run these classes in a Java 8 JVM, you must add -java8 to the list of options passed to the driver.
If you want to Phosphor to cache classes that are dynamically instrumented, then you can add the Java option
-DphosphorCacheDirectory=<CACHE-DIRECTORY> when running your application, where <CACHE-DIRECTORY> is the file
path to the directory where Phosphor should store the cached instrumented class files.
Once you have created an instrumented Java installation (see the section
"Creating an Instrumented Java Installation", you can run your
application with Phosphor.
Locate the JAR for Phosphor's Java agent.
If you have built Phosphor according to the directions described above in the
section "Building Phosphor", then the agent JAR will be available at
Phosphor/target/Phosphor-VERSION.jar relative to the root of this project.
The latest snapshot of the agent JAR is available at the
Sonatype OSS Repository Hosting (OSSRH).
The agent JAR can also be acquired using the Maven dependency:
<build>
...
<dependencies>
<dependency>
<groupId>edu.gmu.swe.phosphor</groupId>
<artifactId>Phosphor</artifactId>
<version>VERSION</version>
</dependency>
</dependencies>
...
</build>Once you have located the JAR for Phosphor's Java agent, execute the same Java command you would normally to run
your application using the Java executable inside the instrumented Java installation you created and
add javaagent and bootclasspath Java options for Phosphor.
If running java using the -jar option run:
<INSTRUMENTED-JAVA-HOME>/bin/java \
-Xbootclasspath/a:<AGENT-JAR> \
-javaagent:<AGENT-JAR> \
[ options ] -jar file.jar [ argument ... ]Otherwise, run:
<INSTRUMENTED-JAVA-HOME>/bin/java \
-Xbootclasspath/a:<AGENT-JAR> \
-javaagent:<AGENT-JAR> \
[options] class [ argument ... ]Where:
Phosphor exposes a simple API to allow data to be labeled with tags, and to retrieve those tags.
Key functionality is implemented in edu.columbia.cs.psl.phosphor.runtime.MultiTainter.
To get or set the taint tag of a primitive type,
developers call the taintedX or getTaint(X) method (replacing X with each of the primitive types, e.g., taintedByte,
etc.).
Ignore the methods ending with the suffix $$PHOSPHOR, they are used internally.
To get or set the taint tag of an object, first cast that object to the interface TaintedWithObjTag (Phosphor changes
all classes to implement this interface), and use the get and set methods.
You can determine if a variable is derived from a particular tainted source by examining the labels on that
variable's Taint object.
You can detaint variables with Phosphor - to do so, use the MultiTainter interface to set the taint on a
value to 0 (or null).
Please note that the control tracking functionality can impose SIGNIFICANT overhead (we've observed > 10x slowdown) depending on the structure of the code you are instrumenting and the amount of tainted data flowing around. This is incredibly un-optimized at this point. This also can make it difficult to apply Phosphor with control tracking to very large methods (since it causes them to grow beyond the maximum size permitted). Nonetheless, we have had great success applying it in various projects --- it works fine on the JDK (perhaps a few internal classes will be too large, but they were not needed in our workloads) and on projects like Tomcat. There are quite a few paths to improving this functionality. If you are interested in helping, please contact us.
Please email Jonathan Bell with comments, suggestions, or questions. This project is still under development, and we welcome any feedback.
This software is released under the MIT license.
Copyright (c) 2013, by The Trustees of Columbia University in the City of New York.
Permission is hereby granted, free of charge, to any person obtaining a copy of this software and associated documentation files (the "Software"), to deal in the Software without restriction, including without limitation the rights to use, copy, modify, merge, publish, distribute, sublicense, and/or sell copies of the Software, and to permit persons to whom the Software is furnished to do so, subject to the following conditions:
The above copyright notice and this permission notice shall be included in all copies or substantial portions of the Software.
THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
Phosphor makes use of the following libraries:
Phosphor's performance tuning is made possible by JProfiler, the java profiler.
The authors of this software are Katherine Hough, Jonathan Bell and Gail Kaiser. Jonathan Bell and Katherine Hough are funded in part by NSF CCF-1763822 and CCF-1844880. Gail Kaiser directs the Programming Systems Laboratory, funded in part by NSF CCF-1161079, NSF CNS-0905246, and NIH U54 CA121852.