This site is the archived OWASP Foundation Wiki and is no longer accepting Account Requests.
To view the new OWASP Foundation website, please visit https://owasp.org

Bytecode obfuscation

From OWASP
Revision as of 11:18, 19 April 2009 by KirstenS (talk | contribs) (How to recover Source Code from Bytecode?)

Jump to: navigation, search

Status

Released: 14/1/2008

Author

Pierre Parrend

Principles

Java is a language where the source code is quite intuitive to read. And in many cases, the compiled bytecode can also be reversed (or decompiled) into source code. This presents problems for projects that require confidentiality of the source code. This article provides an introduction to protecting bytecode through obfuscation.

How to recover Source Code from Bytecode?

There are a number of freely available Java decompilers that all provide similar functionality, including:

  • Recover source code from Java bytecode,
  • Retrieve names of local Variables and parameters,
  • Retrieve comments and JavaDoc

Popular decompilers include:

How to prevent Java code from being Reverse-engineered ?

Several actions can be taken for preventing reverse-engineering :

  • Code Obfuscation. This is done mainly through variable renaming; see next paragraph for more precisions,
  • Suppression of End Of Line Characters. This makes the code difficult to parse,
  • Use of anonymous classes for handling events. This seems not to be handled by many Decompiler; however, JAD copes pretty well with this.
  • Class file encryption. This implies some overhead for uncyphering at runtime. Several tools are available:: Canner, by Cinnabar Systems, or JLock by JSoft. They are available for evaluation, and the first is proposed currently for Windows Platforms only.
  • Replacing the method names with certain characters e.g '/' or '.' in the class header causes the popular decompilation tools such as JAD to dump the source code which is incomprehensible (you cannot determine the control flow from the source).

Note: Beware of 100% Java solutions using encryption to protect class files as these are more than likely snake oil. Since the JVM has to read unencrypted class files at some point, even if the class files are encrypted on the disk, they will have to be decrypted before being passed to the JVM. An attacker could modify the local JVM to simply write the class files to disk in their unencrypted form at this point. (See: Javaworld article).
Conjecture: It's is very easy to circumvent these methods to reveal bytecode using a Java profiler.

What obfuscation tools are available ?

A lot of tools exist for Java code Obfuscation. You can find extensive lists under following URLs, or simply type 'obfuscator' in your favorite search engine :

Among those projects, some are open source project, and therefore more suitable for research - but also for enterprises who wish to control the programs they use (without any warranty):

  • Proguard is a shrinker (make code more compact), and optimizer and obfuscator.
  • Jode is a decompiler, an optimizer and an obfuscator. It contains facilities for cleaning logging statements,,
  • Jarg,
  • Javaguard, which is a simple obfuscator, without many documentation,
  • CafeBabe, which allows precise view of Bytecode files and single file obfuscation; a good tool for teaching ByteCode Structure, more than a production tool.

Using Proguard

The following section provides a short tutorial for using Proguard.

First, download the code under following url and unzip it.

For this tutorial, we use the fr.inria.ares.sfelixutils-0.1.jar package.

Go to the main directory of Proguard. For lauching it, you can use following script with given parameters :

      java -jar lib/proguard.jar @config-genericFrame.pro

config-genericFrame.pro is the option file (do not forget to adapt the libraryjars parameter to your own system) :

-obfuscationdictionary ./examples/dictionaries/compact.txt
-libraryjars /usr/java/j2sdk1.4.2_10/jre/lib/rt.jar
-injars fr.inria.ares.sfelixutils-0.1.jar
-outjar fr.inria.ares.sfelixutils-0.1-obs.jar
-dontshrink
-dontoptimize
-keep public class proguard.ProGuard {
public static void main(java.lang.String[]);
}

Remark that the 'keep' option is mandatory, we use this default class for not keep anything out.

The example dictionnary (here compact.txt) is given with the code.

The output is stored in the package 'genericFrameOut.jar'.

You can observe the modifications implied by obfuscation with following commands :

jar xvf genericFrameOut.jar
cd genericFrame/pub/gui/
jad c.class
more c.jad more c.jad

Remark than Strings are kept unmodified. If you want you code to be hard to read, do not forget to remove any debugging and logging comments. Jode has some facilities for making this easier.

Using CafeBabe

CafeBabe is a convenient tool for teaching structure of ByteCode files. You can download it at this URL.

Unzip it and execute following command :
java -classpath CafeBabe.jar org.javalobby.apps.cafebabe.CafeBabe

Have a look at some class from the original genericFrame.jar package.

Then obfuscate it, and compare both - original and modified class :

  • with the CafeBabe viewer,
  • after decompiling it with JAD.

What conclusion can you draw of it ?

Using Jode

Jode is to be found here with instructions on how to use the decompiler and obfuscator functions here.

Links