Semantic diffusion — sdiff

Volatile-slot masking that resists local value propagation and constant folding by the optimizer.

sdiff spreads a value's computation across a small set of volatile slots, so an optimizer (including an analyst's own -O2) cannot propagate or fold it locally. It complements the algebraic obfuscation of mba with an optimization-resistance property.

Options

Key	Default	Range	Meaning
prob	45	1–100	Probability (%) to apply at a candidate site.
slots	3	1–8	Number of volatile slots used.
maxSites / max / sites	80	1–2000	Maximum transformed sites per function.

Example

__attribute__((annotate("obf: sdiff(prob=35,slots=2,maxSites=60)")))
int step(int acc, int x) { return acc + (x * 31); }

More slots = more diffusion, more cost

Each volatile slot is a genuine memory barrier to the optimizer. Higher slots diffuses the value further but adds real load/store traffic — raise it only for hot, sensitive values.