DragonX/librustzcash
1
0
Fork 0
Code Issues Pull Requests Actions Packages Projects Releases Wiki Activity

General code quality improvements.

Browse Source
This commit is contained in:
Sean Bowe
2018-03-08 00:41:47 -07:00
parent 896b144a7d
commit b6ef12b077
4 changed files with 133 additions and 40 deletions
Download Patch File Download Diff File Expand all files Collapse all files
src/circuit/mod.rs
+77 -18
View File
@@ -29,6 +29,13 @@ use jubjub::{
use constants;
// TODO: This should probably be removed and we
// should use existing helper methods on `Option`
// for mapping with an error.
/// This basically is just an extension to `Option`
/// which allows for a convenient mapping to an
/// error on `None`.
trait Assignment<T> {
fn get(&self) -> Result<&T, SynthesisError>;
}
@@ -42,6 +49,7 @@ impl<T> Assignment<T> for Option<T> {
}
}
/// This is an instance of the `Spend` circuit.
pub struct Spend<'a, E: JubjubEngine> {
pub params: &'a E::Params,
/// Value of the note being spent
@@ -75,6 +83,7 @@ impl<'a, E: JubjubEngine> Circuit<E> for Spend<'a, E> {
)?;
{
// Compute the note value in the exponent
let gv = ecc::fixed_base_multiplication(
cs.namespace(|| "compute the value in the exponent"),
FixedGenerators::ValueCommitmentValue,
@@ -82,12 +91,15 @@ impl<'a, E: JubjubEngine> Circuit<E> for Spend<'a, E> {
self.params
)?;
// Booleanize the randomness
// Booleanize the randomness. This does not ensure
// the bit representation is "in the field" because
// it doesn't matter for security.
let hr = boolean::field_into_boolean_vec_le(
cs.namespace(|| "hr"),
self.value_randomness
)?;
// Compute the randomness in the exponent
let hr = ecc::fixed_base_multiplication(
cs.namespace(|| "computation of randomization for value commitment"),
FixedGenerators::ValueCommitmentRandomness,
@@ -95,12 +107,14 @@ impl<'a, E: JubjubEngine> Circuit<E> for Spend<'a, E> {
self.params
)?;
// Compute the Pedersen commitment to the value
let gvhr = gv.add(
cs.namespace(|| "computation of value commitment"),
&hr,
self.params
)?;
// Expose the commitment as an input to the circuit
gvhr.inputize(cs.namespace(|| "value commitment"))?;
}
@@ -118,6 +132,7 @@ impl<'a, E: JubjubEngine> Circuit<E> for Spend<'a, E> {
// demonstrate the prover knows it. If they know a
// congruency then that's equivalent.
// Compute rk = [rsk] ProvingPublicKey
rk = ecc::fixed_base_multiplication(
cs.namespace(|| "computation of rk"),
FixedGenerators::ProofGenerationKey,
@@ -133,29 +148,40 @@ impl<'a, E: JubjubEngine> Circuit<E> for Spend<'a, E> {
self.params
)?;
// There are no sensible attacks on small order points
// of ak (that we're aware of!) but it's a cheap check,
// so we do it.
ak.assert_not_small_order(
cs.namespace(|| "ak not small order"),
self.params
)?;
// Unpack ak and rk for input to BLAKE2s
// This is the "viewing key" preimage for CRH^ivk
let mut vk = vec![];
let mut rho_preimage = vec![];
vk.extend(
ak.repr(cs.namespace(|| "representation of ak"))?
);
// This is the nullifier randomness preimage for PRF^nr
let mut nr_preimage = vec![];
// Extend vk and nr preimages with the representation of
// rk.
{
let repr_rk = rk.repr(
cs.namespace(|| "representation of rk")
)?;
vk.extend(repr_rk.iter().cloned());
rho_preimage.extend(repr_rk);
nr_preimage.extend(repr_rk);
}
assert_eq!(vk.len(), 512);
assert_eq!(nr_preimage.len(), 256);
// Compute the incoming viewing key
// Compute the incoming viewing key ivk
let mut ivk = blake2s::blake2s(
cs.namespace(|| "computation of ivk"),
&vk,
@@ -164,16 +190,24 @@ impl<'a, E: JubjubEngine> Circuit<E> for Spend<'a, E> {
// Little endian bit order
ivk.reverse();
ivk.truncate(E::Fs::CAPACITY as usize); // drop_5
// Witness g_d
// drop_5 to ensure it's in the field
ivk.truncate(E::Fs::CAPACITY as usize);
// Witness g_d. Ensures the point is on the
// curve, but not its order. If the prover
// manages to witness a commitment in the
// tree, then the Output circuit would have
// already guaranteed this.
// TODO: We might as well just perform the
// check again here, since it's not expensive.
let g_d = ecc::EdwardsPoint::witness(
cs.namespace(|| "witness g_d"),
self.g_d,
self.params
)?;
// Compute pk_d
// Compute pk_d = g_d^ivk
let pk_d = g_d.mul(
cs.namespace(|| "compute pk_d"),
&ivk,
@@ -181,6 +215,7 @@ impl<'a, E: JubjubEngine> Circuit<E> for Spend<'a, E> {
)?;
// Compute note contents
// value (in big endian) followed by g_d and pk_d
let mut note_contents = vec![];
note_contents.extend(value_bits.into_iter().rev());
note_contents.extend(
@@ -206,12 +241,13 @@ impl<'a, E: JubjubEngine> Circuit<E> for Spend<'a, E> {
)?;
{
// Booleanize the randomness
// Booleanize the randomness for the note commitment
let cmr = boolean::field_into_boolean_vec_le(
cs.namespace(|| "cmr"),
self.commitment_randomness
)?;
// Compute the note commitment randomness in the exponent
let cmr = ecc::fixed_base_multiplication(
cs.namespace(|| "computation of commitment randomness"),
FixedGenerators::NoteCommitmentRandomness,
@@ -219,6 +255,8 @@ impl<'a, E: JubjubEngine> Circuit<E> for Spend<'a, E> {
self.params
)?;
// Randomize the note commitment. Pedersen hashes are not
// themselves hiding commitments.
cm = cm.add(
cs.namespace(|| "randomization of note commitment"),
&cmr,
@@ -228,21 +266,30 @@ impl<'a, E: JubjubEngine> Circuit<E> for Spend<'a, E> {
let tree_depth = self.auth_path.len();
// This will store (least significant bit first)
// the position of the note in the tree, for use
// in nullifier computation.
let mut position_bits = vec![];
// Injective encoding.
// This is an injective encoding, as cur is a
// point in the prime order subgroup.
let mut cur = cm.get_x().clone();
for (i, e) in self.auth_path.into_iter().enumerate() {
let cs = &mut cs.namespace(|| format!("merkle tree hash {}", i));
// Determines if the current subtree is the "right" leaf at this
// depth of the tree.
let cur_is_right = boolean::Boolean::from(boolean::AllocatedBit::alloc(
cs.namespace(|| "position bit"),
e.map(|e| e.1)
)?);
// Push this boolean for nullifier computation later
position_bits.push(cur_is_right.clone());
// Witness the authentication path element adjacent
// at this depth.
let path_element = num::AllocatedNum::alloc(
cs.namespace(|| "path element"),
|| {
@@ -250,6 +297,7 @@ impl<'a, E: JubjubEngine> Circuit<E> for Spend<'a, E> {
}
)?;
// Swap the two if the current subtree is on the right
let (xl, xr) = num::AllocatedNum::conditionally_reverse(
cs.namespace(|| "conditional reversal of preimage"),
&cur,
@@ -265,6 +313,7 @@ impl<'a, E: JubjubEngine> Circuit<E> for Spend<'a, E> {
preimage.extend(xl.into_bits_le(cs.namespace(|| "xl into bits"))?);
preimage.extend(xr.into_bits_le(cs.namespace(|| "xr into bits"))?);
// Compute the new subtree value
cur = pedersen_hash::pedersen_hash(
cs.namespace(|| "computation of pedersen hash"),
pedersen_hash::Personalization::MerkleTree(i),
@@ -278,7 +327,10 @@ impl<'a, E: JubjubEngine> Circuit<E> for Spend<'a, E> {
// Expose the anchor
cur.inputize(cs.namespace(|| "anchor"))?;
// Compute the cm + g^position for preventing
// faerie gold attacks
{
// Compute the position in the exponent
let position = ecc::fixed_base_multiplication(
cs.namespace(|| "g^position"),
FixedGenerators::NullifierPosition,
@@ -286,6 +338,7 @@ impl<'a, E: JubjubEngine> Circuit<E> for Spend<'a, E> {
self.params
)?;
// Add the position to the commitment
cm = cm.add(
cs.namespace(|| "faerie gold prevention"),
&position,
@@ -293,30 +346,36 @@ impl<'a, E: JubjubEngine> Circuit<E> for Spend<'a, E> {
)?;
}
// Let's compute rho = BLAKE2s(rk || cm + position)
rho_preimage.extend(
// Let's compute nr = BLAKE2s(rk || cm + position)
nr_preimage.extend(
cm.repr(cs.namespace(|| "representation of cm"))?
);
assert_eq!(rho_preimage.len(), 512);
assert_eq!(nr_preimage.len(), 512);
let mut rho = blake2s::blake2s(
cs.namespace(|| "rho computation"),
&rho_preimage,
// Compute nr
let mut nr = blake2s::blake2s(
cs.namespace(|| "nr computation"),
&nr_preimage,
constants::PRF_NR_PERSONALIZATION
)?;
// Little endian bit order
rho.reverse();
rho.truncate(E::Fs::CAPACITY as usize); // drop_5
nr.reverse();
// We want the randomization in the field to
// simplify outside code.
// TODO: This isn't uniformly random.
nr.truncate(E::Fs::CAPACITY as usize);
// Compute nullifier
let nf = ak.mul(
cs.namespace(|| "computation of nf"),
&rho,
&nr,
self.params
)?;
// Expose the nullifier publicly
nf.inputize(cs.namespace(|| "nullifier"))?;
Ok(())