A computation that provably takes real time — and takes almost no time to check
A Verifiable Delay Function (VDF) is a function that’s deliberately slow to compute — not because of a missing optimization, but by design, using a computation that genuinely cannot be sped up by parallelizing it across more hardware — while producing a proof that lets anyone check the result almost instantly, without redoing the slow part. The term was formalized by Dan Boneh, Joseph Bonneau, Benedikt Bünz, and Ben Fisch at CRYPTO 2018.
The mechanism is simple to state: repeated squaring in a group where the
group’s order is unknown, y = x^(2^T) mod N, T squarings in a row, each one
depending on the last — which is exactly what makes it sequential. Read what
a VDF actually requires for the three properties that make
this precise, or go straight to the construction: the
squaring chain, the group of unknown order, and the Wesolowski proof that
makes verification fast regardless of how long the computation ran.
Try it right here — pick a delay and watch it run, live:
Generating a demo-sized RSA modulus…
N = (512-bit, generated right here in your browser). This is not a trusted setup — this page generated its own p and q, so it knows the group's order and could cheat if it wanted to. A real deployment needs either an honest multi-party trusted-setup ceremony on a much larger modulus, or a class group (no trusted setup needed at all) — see how it works.
y
π (proof)
More on Interactive, including verifying a tampered result to see it get rejected.