From: Rainer Urian on
Hi,
assume we are playing the following ECC-k game:
A chooses a random number s from the range 0..2^k as secret key
B can choose n k-bit elliptic curves and corresponding public points
(ECC_i, P_i, i=1..n). The chossen elliptic curves can be arbitrary but must
be cryptographically strong. (e.g. must be primary, not supersingular ,...)
Then A calculates Q_i = P_i*s in ECC_i for i=1..n and sends the results Q_i
to B.

Now, is it possible for B to guess A's secret key with significant
probability ?

--
Rainer

From: Kristian Gj�steen on
Rainer Urian <rainer(a)urian.eu> wrote:
>assume we are playing the following ECC-k game:
>A chooses a random number s from the range 0..2^k as secret key
>B can choose n k-bit elliptic curves and corresponding public points
>(ECC_i, P_i, i=1..n). The chossen elliptic curves can be arbitrary but must
>be cryptographically strong. (e.g. must be primary, not supersingular ,...)
>Then A calculates Q_i = P_i*s in ECC_i for i=1..n and sends the results Q_i
>to B.
>
>Now, is it possible for B to guess A's secret key with significant
>probability ?

Who knows?

I don't think you can reduce this to something well-studied like EC-DDH
or EC-DLOG.

If you place a few restrictions on the choice of the curves, it's not
known how to embed a trapdoor in an elliptic curve (it seems possible
to embed a practical trapdoor in some very special elliptic curves that
on their own remain cryptographically strong).

One might try techniques similar to the attack on textbook RSA where
someone encrypts the same message to multiple recipients. The idea
would be to choose a number of curves that all lift to the same curve
over the rationals, then use the chinese remainder theorem to lift
the points below to the rational curve, then send them down to curves
where the discrete logarithm is easy, to recover the logarithm of the
entire point. The problem is that the lifted point will have too many
digits to be calculatable. So it wouldn't work.

That's one minute of thinking. Someone else might come up with something
better in 30s.

--
Kristian Gj�steen