I recently learned about code obfuscation. Its nice thing to do, when you have spare time, but I have different question. Why to do it?
First, there are languages in which I am sure its great thing – interpreted ones, like php, JavaScript and much more. There it seems like a good and more secure thing.
Second, there are languages where this seems to have no real effect for me – all the native code compiled languages. Take C for example. when compiled, all the variable names, function names, most of obfuscation techniques go away. If some can make it into native code, it would be things like recursion instead of for cycles and so, but disassembled code will anyway have instead of names some disassembler-generated identifiers, right?
And last category are languages I am not quite sure about. And that’s the main reason I ask. These languages would be Java, C# (.NET),and the last Silverlight used in WP7. I ask because I read some article that state that on WP7 apps, code obfuscation helps preventing code from hacking. But I always thought of byte-code as being very similar to standard assembler codes, therefore again not having any information about real pre-compilation variable names, function names, etc. So, where is the truth?
Do it if you want, but don’t expect any determined person to be scared away by it. There exist de-obfuscators, people can read obfuscated code as well (just as there are people who can read optimized assembly and reconstruct the original C code). Code obfuscation just gives you a false sense of security and might deter a person who is just curious (instead of deterring those who are serious about stealing your code). All it gives you is a false sense of security but no real one. Schneier aptly names this “security theater”.
Yes, many modern languages that retain more information about the source can be obfuscated better than those that are compiled right to machine code. For the latter the compiler already does quite a good job with optimization. Your notion of bytecode being akin to traditional assembler is slightly wrong here, though. Especially .NET bytecode retains enough metadata to reconstruct the original source almost exactly (see Reflector). What isn’t retained there are the names of local variables and arguments to methods. But you still need and retain the method and class names.
Another issue you should be aware of: If you give your customers an obfuscated executable and your program crashes, make sure you have a way of getting the real stacktrace back instead of the obfuscated one. Saying “Sorry, I cannot determine the root cause of why my program killed hours of your work since I chose to obfuscate it” isn’t going to cut it, I guess 🙂