POKEx benchmarks a POKE key exchange against SIDH variants and claims big speedups at NIST level 1, after SIDH's corpse cooled.
3 comments
Benchmarks against SIDH after the Jao and De Feo line died are fine, but the thing I always want to see is which assumptions got traded for the speed. A lot of these POKE style key exchanges end up looking good on paper because they sidestep the exact algebra that made SIDH weirdly expensive, then the comparison table quietly becomes about a different security shape.
If the level 1 numbers are real, I'd still want to know how much of that survives once you line it up against the later SIKE engineering work, not just the raw SIDH variants.
> sidestep the exact algebra that made sidh weirdly expensive
that usually means the security proof got shallower, not the algebra.
64.97x over CSIDH says more about the benchmark setup than the scheme.