4
$\begingroup$

My cryptology professor asked us to show that while the following signature scheme is conceptually valid, that it is inherently insecure, however, I am not sure how to demonstrate that the associated verification process produces an equality if the signature is valid.

The protocol works as follows given:

  • A publicly shared prime $p$;
  • A primitive root of $p$, $a|a<q$
  • A private key: $x|x<q$; and
  • A public key: $y=a^x\bmod{q}$
  • A cryptographic hash function $H=\texttt{SHA1}$

A signature $s$ can be generated for $m$ as follows: first, compute $h = H(m)$; if $\texttt{gcd}(h, q-1) \neq 1$ append $h$ to $m$ and calculate a new hash, and continue the process until $h$ and $q-1$ are relatively prime.

Then, calculate $z|z\cdot h\equiv x\bmod{q-1}$ and return the signature $s = a^{z}$.

In order to verify $s$, a user verifies that $Y=(a^z)^{h}=a^{x}\bmod{q}$.


For example, in order to sign a message $m$, where:

  • $q=6043$
  • $a=5$
  • $x=1098$
  • $y=485$
  • $h\leftarrow H(m)=612515367434372930600221767499032307523881412051$
  • $z\leftarrow z\cdot h\equiv x\bmod{q-1}=8513$
  • $s\leftarrow a^{z} = 5845$

Given the signature $s$, a user can verify that the signature is correct by verifying that $y=(a^{z})^{h}=a^{x}\bmod{q}$, however, I am not sure how to evaluate

$$(a^{z})^{h}= a^{x}\bmod{q}$$

since $h$ is (relatively) large. I (think) that I need to evaluate the following expression, however, I am not sure if its computationally feasible since:

$$(a^{z})^{h}\rightarrow s^{h}\rightarrow 5845^{612515367434372930600221767499032307523881412051}$$

Meaning that in order to verify the signature, I would have to perform the following calculation:

$$(5^{8513})^{612515367434372930600221767499032307523881412051} = 5^{1098}\bmod{6043}$$

Normally, I would use fast modular exponentiation here, but $h$ is absolutely enormous. For a while, I thought that maybe I misunderstood the protocol and $h$ should be reduced modulo $q$, but, again, maybe I just have an embarrassingly flawed understanding of the protocol?

How do you think I should verify the signature?

$\endgroup$
1
$\begingroup$

Sorry, can't comment yet. So I'm going be more verbose since it's an answer ;)

I get other values for $z$ and $s$: $$h^{-1} = 2441\mod q-1 $$ and $$z = x \cdot h^{-1} = 3612\mod q-1\\ \Rightarrow s = a^z = 1039\mod q$$

The hash value $h$ can be reduced since we are working modulo $q$. With $q$ prime we have $\phi(q) = q-1$ and we can reduce the exponent modulo $q-1$.

This gives $$y' = s^h = s^{h\mod q-1} \mod q\\$$ So with $h \equiv 2297 \mod q-1$ we verify: $$y' = 1039^{2297}\mod 6043 = 485 = y$$ as desired.

$\endgroup$
  • $\begingroup$ Sorry, can't upvote yet. Thanks for your response, I never realized that there was an implicit modulo reduction for $h$ in there. $\endgroup$ – hodgepodge Feb 23 '16 at 15:57

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.