The real problem for someone who tried to reverse engineer LFS wouldn't be getting past encryption, or tracing windows api calls.
I would have thought that the big difficulty is that the one thing that you really want - the physics engine - is also by far the most difficult to translate from machine code back to a human readable form.
The problem is that the maths involved are extremely challenging even with fully annotated source code, reams of documents and reference literature
The original source code will be structured in such a way as to clarify which parts of the calculations apply to which attributes of the physics - tyres, aero, temps, wear, forces.. etc. Taking optimized machine code and analysing it to the point where you could seperate these math based components in a way that would allow you to maintain or update them seperately seems like a huge task to me ?
Basically, its not 'machine code --> c++' thats the biggest difficulty, its 'machine code --> math'. Thats a much higher level of abstraction, and there won't be 'WAREZ TOOLZ' to help you with it either