Assume $g$ is generator of multiplicative group modulo prime $p=2q+1$ where $q$ is prime.

Assume we know $g^{2t}\bmod p$ and $g^{2}\bmod p$ and assume we can have access to a Diffie-Hellman oracle.

Can we find $g^t\bmod p$ in polynomial time?

Note if we can do that we can break discrete log with access to a DH oracle when generator order is even.


1 Answer 1


Can we find $g^t \bmod p$ in polynomial time?

We can find either $g^t$ or $-g^{t} = g^{t + (p-1)/2}$; obviously, we can't tell which one was the correct one with the information we were given.

Because $p \equiv 3 \pmod 4$ (because $(p-1)/2$ is assumed to be prime, and taking $p=5$ off the table - that can be handled as a special case), then [1] we can compute modular square-roots with the simple computation $\sqrt{x} = \pm x^{(p+1)/4}$.

So, we have $g^t \in\{ -(g^{2t})^{(p+1)/4},+(g^{2t})^{(p+1)/4}\}$, easily doable in polytime.

[1]: If $p \equiv 1 \bmod 4$, then it is still practical to compute modular squareroots, it's a bit more involved.

  • $\begingroup$ True.. but can the oracle resolve the ambiguity in the sign? $\endgroup$
    – Turbo
    Mar 27 at 18:25
  • $\begingroup$ @Turbo: no, it can't, because both values are possible solutions for $g^t$ $\endgroup$
    – poncho
    Mar 27 at 19:31

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.