Return-Oriented-Programming (ROP) is a common attack method that malware uses to exploit buffer overflows:
http://en.wikipedia.org/wiki/Return-oriented_programming
It relies on code having predictable contents and being in predictable locations.
The compiler often makes potayto-potahto decisions when generating code. By optionally providing a random value to the compiler, it can use that to toss a coin for the decision, making the generated code slightly different - different enough to defeat many ROP attacks.
Perturbations can be:
changing the stack layout of locals
changing the order of register selection
changing the scheduling order of instructions
weights given to loop variables
instruction selection
etc.
Syntax:
-perturb=seed // use seed to guide the compiler's coin toss
-perturb // have the compiler generate its own seed, likely by
// using the clock. -v will cause this value to be printed
default // use a seed value of 0, causing the same behavior the
// compiler has now
Using this can also shake out compiler bugs by "fuzz" testing of different paths through the compiler. It can help isolate stack corruption code bugs by helping find a more reproducible test case.
This switch can be particularly useful for those who are willing to build their apps from source, so that their executable will be different from anybody else's built from the identical source.
It shouldn't be hard to implement.
Comment #1 by robert.schadek — 2024-12-13T18:30:20Z