This commit adds three new cryptosystems to libsecp256k1:
Pedersen commitments are a system for making blinded commitments
to a value. Functionally they work like:
commit_b,v = H(blind_b || value_v),
except they are additively homorphic, e.g.
C(b1, v1) - C(b2, v2) = C(b1 - b2, v1 - v2) and
C(b1, v1) - C(b1, v1) = 0, etc.
The commitments themselves are EC points, serialized as 33 bytes.
In addition to the commit function this implementation includes
utility functions for verifying that a set of commitments sums
to zero, and for picking blinding factors that sum to zero.
If the blinding factors are uniformly random, pedersen commitments
have information theoretic privacy.
Borromean ring signatures are a novel efficient ring signature
construction for AND/OR admissions policies (the code here implements
an AND of ORs, each of any size). This construction requires
32 bytes of signature per pubkey used plus 32 bytes of constant
overhead. With these you can construct signatures like "Given pubkeys
A B C D E F G, the signer knows the discrete logs
satisifying (A || B) & (C || D || E) & (F || G)".
ZK range proofs allow someone to prove a pedersen commitment is in
a particular range (e.g. [0..2^64)) without revealing the specific
value. The construction here is based on the above borromean
ring signature and uses a radix-4 encoding and other optimizations
to maximize efficiency. It also supports encoding proofs with a
non-private base-10 exponent and minimum-value to allow trading
off secrecy for size and speed (or just avoiding wasting space
keeping data private that was already public due to external
constraints).
A proof for a 32-bit mantissa takes 2564 bytes, but 2048 bytes of
this can be used to communicate a private message to a receiver
who shares a secret random seed with the prover.
Also: get rid of precomputed H tables (Pieter Wuille)
7b50483ad789081ba158799e5b94330f62932607 Adds a declassify operation to aid constant-time analysis. (Gregory Maxwell)
34a67c773b0871e5797c7ab506d004e80911f120 Eliminate harmless non-constant time operations on secret data. (Gregory Maxwell)
Pull request description:
There were several places where the code was non-constant time
for invalid secret inputs. These are harmless under sane use
but get in the way of automatic const-time validation.
(Nonce overflow in signing is not addressed, nor is s==0 in signing)
ACKs for top commit:
sipa:
utACK 7b50483ad789081ba158799e5b94330f62932607
real-or-random:
ACK 7b50483ad789081ba158799e5b94330f62932607 I read the code carefully and tested it
jonasnick:
reACK 7b50483ad789081ba158799e5b94330f62932607
Tree-SHA512: 0776c3a86e723d2f97b9b9cb31d0d0e59dfcf308093b3f46fbc859f73f9957f3fa977d03b57727232040368d058701ef107838f9b1ec98f925ec78ddad495c4e
ECDSA signing has a retry loop for the exceptionally unlikely case
that S==0. S is not a secret at this point and this case is so
rare that it will never be observed but branching on it will trip
up tools analysing if the code is constant time with respect to
secrets.
Derandomized ECDSA can also loop on k being zero or overflowing,
and while k is a secret these cases are too rare (1:2^255) to
ever observe and are also of no concern.
This adds a function for marking memory as no-longer-secret and
sets it up for use with the valgrind memcheck constant-time
test.
It's subtle, since it is actually only touched by hashfp (though
we assert it's non-NULL), but give explicit advice in the default
case.
Signed-off-by: Rusty Russell <rusty@rustcorp.com.au>
Before this commit secp256k1_context_randomize called illegal_callback
when called on a context not initialized for signing. This is not
documented. Moreover, it is not desirable because non-signing contexts
may use randomization in the future.
This commit makes secp256k1_context_randomize a noop in this case. This
is safe because the context cannot be used for signing anyway.
This fixes#573 and it fixesrust-bitcoin/rust-secp256k1#82.
