feat: add traits for group digest

Signed-off-by: Michael Lodder <[email protected]>

Author: mikelodder7

elliptic-curve/Cargo.toml

@@ -29,7 +29,7 @@ zeroize = { version = "1", default-features = false }
 
 # optional dependencies
 base64ct = { version = "1", optional = true, default-features = false }
-digest = { version = "0.9", optional = true, default-features = false }
+digest_traits = { version = "0.9", optional = true, default-features = false, package = "digest" }
 ff = { version = "0.11", optional = true, default-features = false }
 group = { version = "0.11", optional = true, default-features = false }
 hex-literal = { version = "0.3", optional = true }
@@ -47,10 +47,10 @@ alloc = ["der/alloc", "sec1/alloc", "zeroize/alloc"] # todo: use weak activation
 arithmetic = ["ff", "group"]
 bits = ["arithmetic", "ff/bits"]
 dev = ["arithmetic", "hex-literal", "pem", "pkcs8"]
+digest = ["digest_traits", "ff", "group"]
 ecdh = ["arithmetic"]
 hazmat = []
 jwk = ["alloc", "base64ct/alloc", "serde", "serde_json", "zeroize/alloc"]
-osswu  = ["ff"]
 pem = ["alloc", "arithmetic", "pem-rfc7468/alloc", "pkcs8", "sec1/pem"]
 pkcs8 = ["sec1/pkcs8"]
 std = ["alloc", "rand_core/std"]

elliptic-curve/src/group_digest.rs

@@ -0,0 +1,63 @@
+use crate::hash2field::{hash_to_field, ExpandMsg, FromOkm};
+use crate::map2curve::MapToCurve;
+use group::cofactor::CofactorGroup;
+
+/// Adds hashing arbitrary byte sequences to a valid group element
+pub trait GroupDigest {
+    /// The field element representation for a group value with multiple elements
+    type FieldElement: FromOkm + Default + Copy;
+    /// The resulting group element
+    type Output: CofactorGroup<Subgroup = Self::Output>
+        + MapToCurve<FieldElement = Self::FieldElement, Output = Self::Output>;
+
+    /// Computes the hash to curve routine according to
+    /// <https://www.ietf.org/archive/id/draft-irtf-cfrg-hash-to-curve-13.html>
+    /// which says
+    /// Uniform encoding from byte strings to points in G.
+    /// That is, the distribution of its output is statistically close
+    /// to uniform in G.
+    /// This function is suitable for most applications requiring a random
+    /// oracle returning points in G assuming a cryptographically secure
+    /// hash function is used.
+    ///
+    /// Examples
+    ///
+    /// Using a fixed size hash function
+    ///
+    /// ```ignore
+    /// let pt = ProjectivePoint::hash_from_bytes::<hash2field::ExpandMsgXmd<sha2::Sha256>>(b"test data", b"CURVE_XMD:SHA-256_SSWU_RO_");
+    /// ```
+    ///
+    /// Using an extendable output function
+    ///
+    /// ```ignore
+    /// let pt = ProjectivePoint::hash_from_bytes::<hash2field::ExpandMsgXof<sha3::Shake256>>(b"test data", b"CURVE_XOF:SHAKE-256_SSWU_RO_");
+    /// ```
+    ///
+    fn hash_from_bytes<X: ExpandMsg>(msg: &[u8], dst: &'static [u8]) -> Self::Output {
+        let mut u = [Self::FieldElement::default(), Self::FieldElement::default()];
+        hash_to_field::<X, _>(msg, dst, &mut u);
+        let q0 = Self::Output::map_to_curve(u[0]);
+        let q1 = Self::Output::map_to_curve(u[1]);
+        // Ideally we could add and then clear cofactor once
+        // thus saving a call but the field elements may not
+        // add properly due to the underlying implementation
+        // which could result in an incorrect subgroup
+        q0.clear_cofactor() + q1.clear_cofactor()
+    }
+
+    /// Computes the encode to curve routine according to
+    /// <https://www.ietf.org/archive/id/draft-irtf-cfrg-hash-to-curve-13.html>
+    /// which says
+    /// Nonuniform encoding from byte strings to
+    /// points in G. That is, the distribution of its output is not
+    /// uniformly random in G: the set of possible outputs of
+    /// encode_to_curve is only a fraction of the points in G, and some
+    /// points in this set are more likely to be output than others.
+    fn encode_from_bytes<X: ExpandMsg>(msg: &[u8], dst: &'static [u8]) -> Self::Output {
+        let mut u = [Self::FieldElement::default()];
+        hash_to_field::<X, _>(msg, dst, &mut u);
+        let q0 = Self::Output::map_to_curve(u[0]);
+        q0.clear_cofactor()
+    }
+}

elliptic-curve/src/hash2field.rs

@@ -2,32 +2,32 @@ mod expand_msg;
 mod expand_msg_xmd;
 mod expand_msg_xof;
 
-use core::convert::TryFrom;
 pub use expand_msg::*;
 pub use expand_msg_xmd::*;
 pub use expand_msg_xof::*;
+use generic_array::{typenum::Unsigned, ArrayLength, GenericArray};
 
 /// The trait for helping to convert to a scalar
-pub trait FromOkm<const L: usize>: Sized {
+pub trait FromOkm {
+    /// The number of bytes needed to convert to a scalar
+    type Length: ArrayLength<u8>;
+
     /// Convert a byte sequence into a scalar
-    fn from_okm(data: &[u8; L]) -> Self;
+    fn from_okm(data: &GenericArray<u8, Self::Length>) -> Self;
 }
 
 /// Convert an arbitrary byte sequence according to
 /// <https://tools.ietf.org/html/draft-irtf-cfrg-hash-to-curve-11#section-5.3>
-pub fn hash_to_field<E, T, const L: usize, const COUNT: usize, const OUT: usize>(
-    data: &[u8],
-    domain: &[u8],
-) -> [T; COUNT]
+pub fn hash_to_field<E, T>(data: &[u8], domain: &'static [u8], out: &mut [T])
 where
-    E: ExpandMsg<OUT>,
-    T: FromOkm<L> + Default + Copy,
+    E: ExpandMsg,
+    T: FromOkm + Default,
 {
-    let random_bytes = E::expand_message(data, domain);
-    let mut out = [T::default(); COUNT];
-    for i in 0..COUNT {
-        let u = <[u8; L]>::try_from(&random_bytes[(L * i)..L * (i + 1)]).expect("not enough bytes");
-        out[i] = T::from_okm(&u);
+    let len_in_bytes = T::Length::to_usize() * out.len();
+    let mut tmp = GenericArray::<u8, <T as FromOkm>::Length>::default();
+    let mut expander = E::expand_message(data, domain, len_in_bytes);
+    for o in out.iter_mut() {
+        expander.fill_bytes(&mut tmp);
+        *o = T::from_okm(&tmp);
     }
-    out
 }

elliptic-curve/src/hash2field/expand_msg.rs

@@ -1,5 +1,74 @@
+use digest_traits::{Digest, ExtendableOutputDirty, Update, XofReader};
+use generic_array::{ArrayLength, GenericArray};
+
+/// Salt when the DST is too long
+const OVERSIZE_DST_SALT: &[u8] = b"H2C-OVERSIZE-DST-";
+/// Maximum domain separation tag length
+const MAX_DST_LEN: usize = 255;
+
 /// Trait for types implementing expand_message interface for hash_to_field
-pub trait ExpandMsg<const OUT: usize> {
-    /// Expands `msg` to the required number of bytes in `buf`
-    fn expand_message(msg: &[u8], dst: &[u8]) -> [u8; OUT];
+pub trait ExpandMsg {
+    /// Expands `msg` to the required number of bytes
+    /// Returns an expander that can be used to call `read` until enough
+    /// bytes have been consumed
+    fn expand_message(msg: &[u8], dst: &'static [u8], len_in_bytes: usize) -> Self;
+
+    /// Fill the array with the expanded bytes
+    fn fill_bytes(&mut self, okm: &mut [u8]);
+}
+
+/// The domain separation tag
+///
+/// Implements [section 5.4.3 of `draft-irtf-cfrg-hash-to-curve-13`][dst].
+///
+/// [dst]: https://datatracker.ietf.org/doc/html/draft-irtf-cfrg-hash-to-curve-13#section-5.4.3
+pub(crate) enum Domain<L: ArrayLength<u8>> {
+    /// > 255
+    Hashed(GenericArray<u8, L>),
+    /// <= 255
+    Array(&'static [u8]),
+}
+
+impl<L: ArrayLength<u8>> Domain<L> {
+    pub fn xof<X>(dst: &'static [u8]) -> Self
+    where
+        X: Default + ExtendableOutputDirty + Update,
+    {
+        if dst.len() > MAX_DST_LEN {
+            let mut data = GenericArray::<u8, L>::default();
+            X::default()
+                .chain(OVERSIZE_DST_SALT)
+                .chain(dst)
+                .finalize_xof_dirty()
+                .read(&mut data);
+            Self::Hashed(data)
+        } else {
+            Self::Array(dst)
+        }
+    }
+
+    pub fn xmd<X>(dst: &'static [u8]) -> Self
+    where
+        X: Digest<OutputSize = L>,
+    {
+        if dst.len() > MAX_DST_LEN {
+            Self::Hashed(X::new().chain(OVERSIZE_DST_SALT).chain(dst).finalize())
+        } else {
+            Self::Array(dst)
+        }
+    }
+
+    pub fn data(&self) -> &[u8] {
+        match self {
+            Self::Hashed(d) => &d[..],
+            Self::Array(d) => *d,
+        }
+    }
+
+    pub fn len(&self) -> usize {
+        match self {
+            Self::Hashed(_) => L::to_usize(),
+            Self::Array(d) => d.len(),
+        }
+    }
 }

elliptic-curve/src/hash2field/expand_msg_xmd.rs

@@ -1,72 +1,94 @@
-use super::ExpandMsg;
-use core::marker::PhantomData;
-use digest::{
+use super::{Domain, ExpandMsg};
+use digest_traits::{
     generic_array::{typenum::Unsigned, GenericArray},
     BlockInput, Digest,
 };
-use subtle::{Choice, ConditionallySelectable};
 
 /// Placeholder type for implementing expand_message_xmd based on a hash function
-#[derive(Debug)]
-pub struct ExpandMsgXmd<HashT> {
-    phantom: PhantomData<HashT>,
+pub struct ExpandMsgXmd<HashT>
+where
+    HashT: Digest + BlockInput,
+{
+    b_0: GenericArray<u8, HashT::OutputSize>,
+    b_vals: GenericArray<u8, HashT::OutputSize>,
+    domain: Domain<HashT::OutputSize>,
+    index: usize,
+    offset: usize,
+    ell: usize,
+}
+
+impl<HashT> ExpandMsgXmd<HashT>
+where
+    HashT: Digest + BlockInput,
+{
+    fn next(&mut self) -> bool {
+        if self.index < self.ell {
+            self.index += 1;
+            self.offset = 0;
+            // b_0 XOR b_(idx - 1)
+            let mut tmp = GenericArray::<u8, HashT::OutputSize>::default();
+            self.b_0
+                .iter()
+                .zip(&self.b_vals[..])
+                .enumerate()
+                .for_each(|(j, (b0val, bi1val))| tmp[j] = b0val ^ bi1val);
+            self.b_vals = HashT::new()
+                .chain(tmp)
+                .chain([self.index as u8])
+                .chain(self.domain.data())
+                .chain([self.domain.len() as u8])
+                .finalize();
+            true
+        } else {
+            false
+        }
+    }
 }
 
 /// ExpandMsgXmd implements expand_message_xmd for the ExpandMsg trait
-impl<HashT, const LEN_IN_BYTES: usize> ExpandMsg<LEN_IN_BYTES> for ExpandMsgXmd<HashT>
+impl<HashT> ExpandMsg for ExpandMsgXmd<HashT>
 where
     HashT: Digest + BlockInput,
 {
-    fn expand_message(msg: &[u8], dst: &[u8]) -> [u8; LEN_IN_BYTES] {
+    fn expand_message(msg: &[u8], dst: &'static [u8], len_in_bytes: usize) -> Self {
         let b_in_bytes = HashT::OutputSize::to_usize();
-        let ell = (LEN_IN_BYTES + b_in_bytes - 1) / b_in_bytes;
+        let ell = (len_in_bytes + b_in_bytes - 1) / b_in_bytes;
         if ell > 255 {
             panic!("ell was too big in expand_message_xmd");
         }
+        let domain = Domain::xmd::<HashT>(dst);
         let b_0 = HashT::new()
             .chain(GenericArray::<u8, HashT::BlockSize>::default())
             .chain(msg)
-            .chain([(LEN_IN_BYTES >> 8) as u8, LEN_IN_BYTES as u8, 0u8])
-            .chain(dst)
-            .chain([dst.len() as u8])
+            .chain([(len_in_bytes >> 8) as u8, len_in_bytes as u8, 0u8])
+            .chain(domain.data())
+            .chain([domain.len() as u8])
             .finalize();
 
-        let mut b_vals = HashT::new()
+        let b_vals = HashT::new()
             .chain(&b_0[..])
             .chain([1u8])
-            .chain(dst)
-            .chain([dst.len() as u8])
+            .chain(domain.data())
+            .chain([domain.len() as u8])
             .finalize();
 
-        let mut buf = [0u8; LEN_IN_BYTES];
-        let mut offset = 0;
+        Self {
+            b_0,
+            b_vals,
+            domain,
+            index: 1,
+            offset: 0,
+            ell,
+        }
+    }
 
-        for i in 1..ell {
-            // b_0 XOR b_(idx - 1)
-            let mut tmp = GenericArray::<u8, HashT::OutputSize>::default();
-            b_0.iter()
-                .zip(&b_vals[..])
-                .enumerate()
-                .for_each(|(j, (b0val, bi1val))| tmp[j] = b0val ^ bi1val);
-            for b in b_vals {
-                buf[offset % LEN_IN_BYTES].conditional_assign(
-                    &b,
-                    Choice::from(if offset < LEN_IN_BYTES { 1 } else { 0 }),
-                );
-                offset += 1;
+    fn fill_bytes(&mut self, okm: &mut [u8]) {
+        for b in okm {
+            if self.offset == self.b_vals.len() && !self.next() {
+                return;
             }
-            b_vals = HashT::new()
-                .chain(tmp)
-                .chain([(i + 1) as u8])
-                .chain(dst)
-                .chain([dst.len() as u8])
-                .finalize();
-        }
-        for b in b_vals {
-            buf[offset % LEN_IN_BYTES]
-                .conditional_assign(&b, Choice::from(if offset < LEN_IN_BYTES { 1 } else { 0 }));
-            offset += 1;
+            *b = self.b_vals[self.offset];
+            self.offset += 1;
         }
-        buf
     }
 }

elliptic-curve/src/hash2field/expand_msg_xof.rs

@@ -1,27 +1,33 @@
 use super::ExpandMsg;
-use core::marker::PhantomData;
-use digest::{ExtendableOutput, Update, XofReader};
+use crate::hash2field::Domain;
+use digest_traits::{ExtendableOutput, ExtendableOutputDirty, Update, XofReader};
+use generic_array::typenum::U32;
 
 /// Placeholder type for implementing expand_message_xof based on an extendable output function
-#[derive(Debug)]
-pub struct ExpandMsgXof<HashT> {
-    phantom: PhantomData<HashT>,
+pub struct ExpandMsgXof<HashT>
+where
+    HashT: Default + ExtendableOutput + ExtendableOutputDirty + Update,
+{
+    reader: <HashT as ExtendableOutput>::Reader,
 }
 
 /// ExpandMsgXof implements expand_message_xof for the ExpandMsg trait
-impl<HashT, const LEN_IN_BYTES: usize> ExpandMsg<LEN_IN_BYTES> for ExpandMsgXof<HashT>
+impl<HashT> ExpandMsg for ExpandMsgXof<HashT>
 where
-    HashT: Default + ExtendableOutput + Update,
+    HashT: Default + ExtendableOutput + ExtendableOutputDirty + Update,
 {
-    fn expand_message(msg: &[u8], dst: &[u8]) -> [u8; LEN_IN_BYTES] {
-        let mut buf = [0u8; LEN_IN_BYTES];
-        let mut r = HashT::default()
+    fn expand_message(msg: &[u8], dst: &'static [u8], len_in_bytes: usize) -> Self {
+        let domain = Domain::<U32>::xof::<HashT>(dst);
+        let reader = HashT::default()
             .chain(msg)
-            .chain([(LEN_IN_BYTES >> 8) as u8, LEN_IN_BYTES as u8])
-            .chain(dst)
-            .chain([dst.len() as u8])
+            .chain([(len_in_bytes >> 8) as u8, len_in_bytes as u8])
+            .chain(domain.data())
+            .chain([domain.len() as u8])
             .finalize_xof();
-        r.read(&mut buf);
-        buf
+        Self { reader }
+    }
+
+    fn fill_bytes(&mut self, okm: &mut [u8]) {
+        self.reader.read(okm);
     }
 }