[LLVMdev] Re: Hide visible string in variable (Chris Lattner)

[Zhang Qiuyu]

Hi,

Thanks so much at first. 

> Here are some observations:
> 
> > for C level,
> >
> > char a[]="global string test";
> > for(i=0;i<strlen(a);i++){
> >    a[i]=  a[i]^RANDMON;
> > }
> 
> If you compile this C code, "global string test" will occur in the program
> binary, so you have not obfuscated anything.  You can construct exactly
> what you have above in LLVM (just write it as C, compile it to LLVM and
> you'll see what you need to generate), but I don't think this is what you
> want.  What you really want is:
> 
> char a[]="GLOBAL STRING TEST";
> 
> and when the program starts up (perhaps in main), you want to insert this:
> 
> for (i=0...)
>   A[i] = tolower(a[i]);
> 
> Note that you can use whatever function you wanted, obviously uppercasing
> the string isn't much obfuscation.
> 
> To do this on LLVM, you have to do these things:
> 
> 1. Read the string data as a constant (It's an instance of ConstantArray,
>    which you get form the Globalvaraible with getInitializer() as you are
>    doing.

Following your suggestion, I got some progress. Thanks again. But I am still stuck in some problems.

    Constant *Cstr = GV->getInitializer();

After that, I tried to use
    a. 
        for(unsigned i=0;i<Cstr->getNumOperands();++i){
            Cstr->getOperand(i);
       }

    b. for(User::op_iterator I=Cstr->op_begin(),E=Cstr->op_end(); I!=E;++I){
        
            std::cerr<<*I;
    }

   From either a or b, I could get each element of Global Variable. Supposedly, I will use my arithmetic like XOR etc to encode/hide the string. But I cannot use XOR, I mean I tried (*I)^0x33, it doesn't work. I tried op_xor, but I don't know how to use it. For C level, it is really staightforward. But here, I don't know how to do it. It should be easy way to do it. But I spent several hours on it. For simplifing prolem, I also tried to do the way like 
        a[i]=a[i]+1;
but I failed. Shy.

> 2. Construct a new ConstantArray with all of the elements of the original
>    string, but modified according to the function you want (exclusive or
>    is a reasonable start).

    As you said, how could construct a new ConstantArray? Is it like 
    Constant *pC = new Constant(SBtype); // Constant(const Type *Ty)    

    for ConstantArray, replaceUsesOfWithOnConstant(Value *From, Value *To,
                                           bool DisableChecking = false); this
    API replaceUsesOfWithOnConstant seems to be able to do what I want, but how could create/construct a new Value with my owner value?

for  3,4,5, those should work well.

I would really appreciated if you can give me a very simple example to show me how to do it. 

> 3. Change the initializer of the global variable to the new constant with
>    setInitializer().
> 4. Clear the "constant" flag on the string, because the program will be
>    dynamically hacking on the string: GV->setConstant(false);
> 5. Insert the for loop that translates the string when main runs.
> 
> For #5, write the for loop you want, compile it with llvmgcc, then figure
> out how to generate it at compile time.  Alternatively, you could put the
> 'decryption' routine in a library and just insert a call to the library.
> 
> -Chris
>