frost-core 3.0.0

//! Non-adjacent form (NAF) implementations for fast batch scalar multiplication

// We expect slicings in this module to never panic due to algorithmic
// constraints.
#![allow(clippy::indexing_slicing)]

use core::{
    borrow::Borrow,
    fmt::{Debug, Result},
    marker::PhantomData,
};

use alloc::vec::Vec;

use crate::{Ciphersuite, Element, Field, Group, Scalar};

/// A trait for transforming a scalar generic over a ciphersuite to a non-adjacent form (NAF).
pub trait NonAdjacentForm<C: Ciphersuite> {
    fn non_adjacent_form(&self, w: usize) -> Vec<i8>;
}

impl<C> NonAdjacentForm<C> for Scalar<C>
where
    C: Ciphersuite,
{
    /// Computes a width-(w) "Non-Adjacent Form" of this scalar.
    ///
    /// Thanks to curve25519-dalek for the original implementation that informed this one.
    ///
    /// # Safety
    ///
    /// The full scalar field MUST fit in 256 bits in this implementation.
    fn non_adjacent_form(&self, w: usize) -> Vec<i8> {
        // required by the NAF definition
        debug_assert!(w >= 2);
        // required so that the NAF digits fit in i8
        debug_assert!(w <= 8);

        use byteorder::{ByteOrder, LittleEndian};

        let serialized_scalar = <<C::Group as Group>::Field>::little_endian_serialize(self);
        // The canonical serialization length of this `Scalar` in bytes.
        let serialization_len = serialized_scalar.as_ref().len();

        // Compute the size of the non-adjacent form from the number of bytes needed to serialize
        // `Scalar`s, plus 1 bit.
        //
        // The length of the NAF is at most one more than the bit length.
        let naf_length: usize = serialization_len * u8::BITS as usize + 1;

        // Safety:
        //
        // The max value of `naf_length` (the number of bits to represent the
        // scalar plus 1) _should_ have plenty of room in systems where usize is
        // greater than 8 bits (aka, not a u8). If you are able to compile this
        // code on a system with 8-bit pointers, well done, but this code will
        // probably not compute the right thing for you, use a 16-bit or above
        // system. Since the rest of this code uses u64's for limbs, we
        // recommend a 64-bit system.
        let mut naf = vec![0; naf_length];

        // Get the number of 64-bit limbs we need.
        let num_limbs: usize = naf_length.div_ceil(u64::BITS as usize);

        let mut x_u64 = vec![0u64; num_limbs];

        // This length needs to be 8*destination.len(), so pad out to length num_limbs * 8.
        let mut padded_le_serialized = vec![0u8; num_limbs * 8];

        padded_le_serialized[..serialization_len].copy_from_slice(serialized_scalar.as_ref());

        LittleEndian::read_u64_into(padded_le_serialized.as_ref(), &mut x_u64[0..num_limbs]);

        let width = 1 << w;
        let window_mask = width - 1;

        let mut pos = 0;
        let mut carry = 0;
        while pos < naf_length {
            // Construct a buffer of bits of the scalar, starting at bit `pos`
            let u64_idx = pos / 64;
            let bit_idx = pos % 64;

            let bit_buf: u64 = if bit_idx < 64 - w {
                // This window's bits are contained in a single u64
                x_u64[u64_idx] >> bit_idx
            } else {
                // Combine the current u64's bits with the bits from the next u64
                (x_u64[u64_idx] >> bit_idx) | (x_u64[1 + u64_idx] << (64 - bit_idx))
            };

            // Add the carry into the current window
            let window = carry + (bit_buf & window_mask);

            if window & 1 == 0 {
                // If the window value is even, preserve the carry and continue.
                // Why is the carry preserved?
                // If carry == 0 and window & 1 == 0, then the next carry should be 0
                // If carry == 1 and window & 1 == 0, then bit_buf & 1 == 1 so the next carry should be 1
                pos += 1;
                continue;
            }

            if window < width / 2 {
                carry = 0;
                naf[pos] = window as i8;
            } else {
                carry = 1;
                naf[pos] = (window as i8).wrapping_sub(width as i8);
            }

            pos += w;
        }

        naf
    }
}

/// A trait for variable-time multiscalar multiplication without precomputation.
///
/// Implement for a group element.
pub trait VartimeMultiscalarMul<C: Ciphersuite>: Clone {
    /// Given an iterator of public scalars and an iterator of
    /// `Option`s of group elements, compute either `Some(Q)`, where
    /// $$
    /// Q = c\_1 E\_1 + \cdots + c\_n E\_n,
    /// $$
    /// if all points were `Some(E_i)`, or else return `None`.
    fn optional_multiscalar_mul<I, J>(scalars: I, elements: J) -> Option<Self>
    where
        I: IntoIterator,
        I::Item: Borrow<Scalar<C>>,
        J: IntoIterator<Item = Option<Self>>;

    /// Given an iterator of public scalars and an iterator of
    /// public group elements, compute
    /// $$
    /// Q = c\_1 E\_1 + \cdots + c\_n E\_n,
    /// $$
    /// using variable-time operations.
    ///
    /// It is an error to call this function with two iterators of different lengths.
    fn vartime_multiscalar_mul<I, J>(scalars: I, elements: J) -> Self
    where
        I: IntoIterator,
        I::Item: Borrow<Scalar<C>>,
        J: IntoIterator,
        J::Item: Borrow<Self>,
    {
        Self::optional_multiscalar_mul(
            scalars,
            elements.into_iter().map(|e| Some(e.borrow().clone())),
        )
        .expect("all elements should be Some")
    }
}

impl<C> VartimeMultiscalarMul<C> for Element<C>
where
    C: Ciphersuite,
{
    #[allow(clippy::comparison_chain)]
    fn optional_multiscalar_mul<I, J>(scalars: I, elements: J) -> Option<Element<C>>
    where
        I: IntoIterator,
        I::Item: Borrow<Scalar<C>>,
        J: IntoIterator<Item = Option<Element<C>>>,
    {
        let nafs: Vec<_> = scalars
            .into_iter()
            .map(|c| NonAdjacentForm::<C>::non_adjacent_form(c.borrow(), 5))
            .collect();

        let lookup_tables = elements
            .into_iter()
            .map(|P_opt| P_opt.map(|P| LookupTable5::<C, Element<C>>::from(&P)))
            .collect::<Option<Vec<_>>>()?;

        if nafs.len() != lookup_tables.len() {
            return None;
        }

        let mut r = <C::Group>::identity();

        // All NAFs will have the same size, so get it from the first
        if nafs.is_empty() {
            return Some(r);
        }
        let naf_length = nafs[0].len();

        for i in (0..naf_length).rev() {
            let mut t = r + r;

            for (naf, lookup_table) in nafs.iter().zip(lookup_tables.iter()) {
                if naf[i] > 0 {
                    t = t + lookup_table.select(naf[i] as usize);
                } else if naf[i] < 0 {
                    t = t - lookup_table.select(-naf[i] as usize);
                }
            }

            r = t;
        }

        Some(r)
    }
}

/// Holds odd multiples 1A, 3A, ..., 15A of a point A.
#[derive(Copy, Clone)]
pub(crate) struct LookupTable5<C, T> {
    pub(crate) bytes: [T; 8],
    pub(crate) _marker: PhantomData<C>,
}

impl<C: Ciphersuite, T: Copy> LookupTable5<C, T> {
    /// Given public, odd \\( x \\) with \\( 0 < x < 2^4 \\), return \\(xA\\).
    pub fn select(&self, x: usize) -> T {
        debug_assert_eq!(x & 1, 1);
        debug_assert!(x < 16);

        self.bytes[x / 2]
    }
}

impl<C: Ciphersuite, T: Debug> Debug for LookupTable5<C, T> {
    fn fmt(&self, f: &mut core::fmt::Formatter) -> Result {
        write!(f, "LookupTable5({:?})", self.bytes)
    }
}

impl<'a, C> From<&'a Element<C>> for LookupTable5<C, Element<C>>
where
    C: Ciphersuite,
{
    fn from(A: &'a Element<C>) -> Self {
        let mut Ai = [*A; 8];
        let A2 = *A + *A;
        for i in 0..7 {
            Ai[i + 1] = A2 + Ai[i];
        }

        // Now Ai = [A, 3A, 5A, 7A, 9A, 11A, 13A, 15A]
        LookupTable5 {
            bytes: Ai,
            _marker: PhantomData,
        }
    }
}