Reference on Dalvik or Java Virtual Machines?

I am looking into Dalvik bytecode currently but as I lack a compiler background, I am finding it 开发者_JS百科a little hard to grasp the design. I am pretty sure no one has written a book on Dalvik (or I could be wrong) so can someone suggest me a reference on the Java VM that contains some hands-on examples? Specifically, what I am interested in are:

• Understand how to interpret the generated byte-code
• Using VM specifications (Dalvik or Java) to decompile the byte code into an intermediate representation and then compile it back

In short, probably what I am looking for is to learn reverse engineering byte code so that I can analyze it for vulnerabilities. Any suggestions?

---

For reference material, nothing beats the dalvik docs. You can find these either in the dalvik sub-project in AOSP, or they are now available online at http://s.android.com/tech/dalvik/index.html

Bytecode format (or dalvik-bytecode.html in the dalvik project) is probably the one that you would be most interested in. .Dex Format (dex-format.html) is also useful, as is Instruction Formats (instruction-formats.html)

For some more general information about the bytecode, check out http://code.google.com/p/smali/wiki/Registers and http://code.google.com/p/smali/wiki/TypesMethodsAndFields

You'll definitely want a few tools. I'm naturally quite partial to smali/baksmali, which is the only assembler/disassembler pair currently available. There is also a disassembler called dedexer (but no assembler), and dexdump, which comes with the AOSP codebase and provides a low-level dump of dex files - not just the bytecode, but all the dex structures as well (baksmali has similar output, with the -D option).

You might also be interested in apktool, which uses smali/baksmali, but also has the ability to reverse the "compiled" xml files in an apk.

There are a couple of tools out there that convert dalvik bytecode back to java bytecode, although I don't think they're 100% yet - undx and dex2jar

---

There are already tools for reverse engineering .dex files to generate a human-readable representations of the byte codes. One of the most popular is baksmali, which you can find here: http://code.google.com/p/smali/.

A description of the byte codes themselves can be found easily by Googling. Here was the third results: http://www.netmite.com/android/mydroid/dalvik/docs/dalvik-bytecode.html.

If you are reverse engineering layouts, you'll need a binary-xml-to-xml converter as well. There's another stack overflow question that mentions a few tools for doing that: Parse versionCode from android apk files