# In a additive group is it hard to calculate $bg$ given $ag, g, abg$

The ECDH problem defined that given $g,ag,bg$ it is difficult to calculate $abg$. But it is also difficult to calculate $bg$ given $ag,g,abg$. where $g$ is generator and a,b are elements of group.

• Can you explain the specific relationship to cryptography? Otherwise this should be on a different site. Likely Math or Computer Science. Feb 23, 2015 at 13:19
• Note that $a$ and $b$ are normally not elements of the group, they are integers (modulo the order of $g$ if one wants to be pedantic). Feb 23, 2015 at 15:19
• @fkraiem: Exactly, this is what I was just observing. Feb 23, 2015 at 15:20
• yes you are correct a,b are integers not element of group.
– Aria
Feb 24, 2015 at 5:12

$$g' = ag$$ $$a' = a^{-1}$$ $$b' = b$$
Then the restatement of your problem is: given $g' = ag$, $a'g' = g$ and $b'g' = abg$, compute $a'b'g' = bg$, which is exactly the ECDH problem.
Now, this assumes that $a$ has an inverse; this is not a problem if the curve order is a prime, and is easy to work around if the curve order is a prime times a small cofactor.
• I think that the problem can't be solved in a general way. Assume if we can define on this group a pairing e. then chosing $x_0$ in such a way that $e(g,x_0)=\alpha$ then if the DLP is easy in this group, then we can solve for any integer a (modulo the order) $e(a.g, x_0)=\alpha^a$. It depend on the nature of the group. Feb 23, 2015 at 20:45