1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 192 193 194 195 196 197 198 199 200 201 202 203 204 205 206 207 208 209 210 211 212 213 214 215 216 217 218 219 220 221 222 223 224 225 226 227 228 229 230 231 232 233 234 235 236 237 238 239 240
//! From libsecp256k1: //! //! The Secp256k1 curve has an endomorphism, where lambda * (x, y) = (beta * x, y), where //! lambda is {0x53,0x63,0xad,0x4c,0xc0,0x5c,0x30,0xe0,0xa5,0x26,0x1c,0x02,0x88,0x12,0x64,0x5a, //! 0x12,0x2e,0x22,0xea,0x20,0x81,0x66,0x78,0xdf,0x02,0x96,0x7c,0x1b,0x23,0xbd,0x72} //! //! "Guide to Elliptic Curve Cryptography" (Hankerson, Menezes, Vanstone) gives an algorithm //! (algorithm 3.74) to find k1 and k2 given k, such that k1 + k2 * lambda == k mod n, and k1 //! and k2 have a small size. //! It relies on constants a1, b1, a2, b2. These constants for the value of lambda above are: //! //! - a1 = {0x30,0x86,0xd2,0x21,0xa7,0xd4,0x6b,0xcd,0xe8,0x6c,0x90,0xe4,0x92,0x84,0xeb,0x15} //! - b1 = -{0xe4,0x43,0x7e,0xd6,0x01,0x0e,0x88,0x28,0x6f,0x54,0x7f,0xa9,0x0a,0xbf,0xe4,0xc3} //! - a2 = {0x01,0x14,0xca,0x50,0xf7,0xa8,0xe2,0xf3,0xf6,0x57,0xc1,0x10,0x8d,0x9d,0x44,0xcf,0xd8} //! - b2 = {0x30,0x86,0xd2,0x21,0xa7,0xd4,0x6b,0xcd,0xe8,0x6c,0x90,0xe4,0x92,0x84,0xeb,0x15} //! //! The algorithm then computes c1 = round(b1 * k / n) and c2 = round(b2 * k / n), and gives //! k1 = k - (c1*a1 + c2*a2) and k2 = -(c1*b1 + c2*b2). Instead, we use modular arithmetic, and //! compute k1 as k - k2 * lambda, avoiding the need for constants a1 and a2. //! //! g1, g2 are precomputed constants used to replace division with a rounded multiplication //! when decomposing the scalar for an endomorphism-based point multiplication. //! //! The possibility of using precomputed estimates is mentioned in "Guide to Elliptic Curve //! Cryptography" (Hankerson, Menezes, Vanstone) in section 3.5. //! //! The derivation is described in the paper "Efficient Software Implementation of Public-Key //! Cryptography on Sensor Networks Using the MSP430X Microcontroller" (Gouvea, Oliveira, Lopez), //! Section 4.3 (here we use a somewhat higher-precision estimate): //! d = a1*b2 - b1*a2 //! g1 = round((2^272)*b2/d) //! g2 = round((2^272)*b1/d) //! //! (Note that 'd' is also equal to the curve order here because [a1,b1] and [a2,b2] are found //! as outputs of the Extended Euclidean Algorithm on inputs 'order' and 'lambda'). //! //! @fjarri: //! //! To be precise, the method used here is based on "An Alternate Decomposition of an Integer for //! Faster Point Multiplication on Certain Elliptic Curves" by Young-Ho Park, Sangtae Jeong, //! Chang Han Kim, and Jongin Lim: //! <https://link.springer.com/chapter/10.1007%2F3-540-45664-3_23> //! //! The precision used for `g1` and `g2` is not enough to ensure correct approximation at all times. //! For example, `2^272 * b1 / n` used to calculate `g2` is rounded down. //! This means that the approximation `z' = k * g2 / 2^272` always slightly underestimates //! the real value `z = b1 * k / n`. Therefore, when the fractional part of `z` is just slightly //! above 0.5, it will be rounded up, but `z'` will have the fractional part slightly below 0.5 and //! will be rounded down. //! //! The difference `z - z' = k * delta / 2^272`, where `delta = b1 * 2^272 mod n`. //! The closest `z` can get to the fractional part equal to .5 is `1 / (2n)` (since `n` is odd). //! Therefore, to guarantee that `z'` will always be rounded to the same value, one must have //! `delta / 2^m < 1 / (2n * (n - 1))`, where `m` is the power of 2 used for the approximation. //! This means that one should use at least `m = 512` (since `0 < delta < 1`). //! Indeed, tests show that with only `m = 272` the approximation produces off-by-1 errors //! occasionally. //! //! Now since `r1` is calculated as `k - r2 * lambda mod n`, the contract //! `r1 + r2 * lambda = k mod n` is always satisfied. The method guarantees both `r1` and `r2` to be //! less than `sqrt(n)` (so, fit in 128 bits) if the rounding is applied correctly - but in our case //! the off-by-1 errors will produce different `r1` and `r2` which are not necessarily bounded by //! `sqrt(n)`. //! //! In experiments, I was not able to detect any case where they would go outside the 128 bit bound, //! but I cannot be sure that it cannot happen. use crate::arithmetic::{scalar::Scalar, ProjectivePoint}; use core::ops::{Mul, MulAssign}; use elliptic_curve::subtle::{Choice, ConditionallySelectable, ConstantTimeEq}; /// Lookup table containing precomputed values `[p, 2p, 3p, ..., 8p]` struct LookupTable([ProjectivePoint; 8]); impl From<&ProjectivePoint> for LookupTable { fn from(p: &ProjectivePoint) -> Self { let mut points = [*p; 8]; for j in 0..7 { points[j + 1] = p + &points[j]; } LookupTable(points) } } impl LookupTable { /// Given -8 <= x <= 8, returns x * p in constant time. pub fn select(&self, x: i8) -> ProjectivePoint { debug_assert!(x >= -8); debug_assert!(x <= 8); // Compute xabs = |x| let xmask = x >> 7; let xabs = (x + xmask) ^ xmask; // Get an array element in constant time let mut t = ProjectivePoint::identity(); for j in 1..9 { let c = (xabs as u8).ct_eq(&(j as u8)); t.conditional_assign(&self.0[j - 1], c); } // Now t == |x| * p. let neg_mask = Choice::from((xmask & 1) as u8); t.conditional_assign(&-t, neg_mask); // Now t == x * p. t } } const MINUS_LAMBDA: Scalar = Scalar::from_bytes_unchecked(&[ 0xac, 0x9c, 0x52, 0xb3, 0x3f, 0xa3, 0xcf, 0x1f, 0x5a, 0xd9, 0xe3, 0xfd, 0x77, 0xed, 0x9b, 0xa4, 0xa8, 0x80, 0xb9, 0xfc, 0x8e, 0xc7, 0x39, 0xc2, 0xe0, 0xcf, 0xc8, 0x10, 0xb5, 0x12, 0x83, 0xcf, ]); const MINUS_B1: Scalar = Scalar::from_bytes_unchecked(&[ 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0xe4, 0x43, 0x7e, 0xd6, 0x01, 0x0e, 0x88, 0x28, 0x6f, 0x54, 0x7f, 0xa9, 0x0a, 0xbf, 0xe4, 0xc3, ]); const MINUS_B2: Scalar = Scalar::from_bytes_unchecked(&[ 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xfe, 0x8a, 0x28, 0x0a, 0xc5, 0x07, 0x74, 0x34, 0x6d, 0xd7, 0x65, 0xcd, 0xa8, 0x3d, 0xb1, 0x56, 0x2c, ]); const G1: Scalar = Scalar::from_bytes_unchecked(&[ 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x30, 0x86, 0xd2, 0x21, 0xa7, 0xd4, 0x6b, 0xcd, 0xe8, 0x6c, 0x90, 0xe4, 0x92, 0x84, 0xeb, 0x15, 0x3d, 0xab, ]); const G2: Scalar = Scalar::from_bytes_unchecked(&[ 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0xe4, 0x43, 0x7e, 0xd6, 0x01, 0x0e, 0x88, 0x28, 0x6f, 0x54, 0x7f, 0xa9, 0x0a, 0xbf, 0xe4, 0xc4, 0x22, 0x12, ]); /// Find r1 and r2 given k, such that r1 + r2 * lambda == k mod n. fn decompose_scalar(k: &Scalar) -> (Scalar, Scalar) { // these _var calls are constant time since the shift amount is constant let c1 = k.mul_shift_var(&G1, 272); let c2 = k.mul_shift_var(&G2, 272); let c1 = c1 * MINUS_B1; let c2 = c2 * MINUS_B2; let r2 = c1 + c2; let r1 = k + r2 * MINUS_LAMBDA; (r1, r2) } /// Returns `[a_0, ..., a_32]` such that `sum(a_j * 2^(j * 4)) == x`, /// and `-8 <= a_j <= 7`. /// Assumes `x < 2^128`. fn to_radix_16_half(x: &Scalar) -> [i8; 33] { // `x` can have up to 256 bits, so we need an additional byte to store the carry. let mut output = [0i8; 33]; // Step 1: change radix. // Convert from radix 256 (bytes) to radix 16 (nibbles) let bytes = x.to_bytes(); for i in 0..16 { output[2 * i] = (bytes[31 - i] & 0xf) as i8; output[2 * i + 1] = ((bytes[31 - i] >> 4) & 0xf) as i8; } debug_assert!((x >> 128).is_zero().unwrap_u8() == 1); // Step 2: recenter coefficients from [0,16) to [-8,8) for i in 0..32 { let carry = (output[i] + 8) >> 4; output[i] -= carry << 4; output[i + 1] += carry; } output } fn mul_windowed(x: &ProjectivePoint, k: &Scalar) -> ProjectivePoint { let (r1, r2) = decompose_scalar(k); let x_beta = x.endomorphism(); let r1_sign = r1.is_high(); let r1_c = Scalar::conditional_select(&r1, &-r1, r1_sign); let r2_sign = r2.is_high(); let r2_c = Scalar::conditional_select(&r2, &-r2, r2_sign); let table1 = LookupTable::from(&ProjectivePoint::conditional_select(x, &-x, r1_sign)); let table2 = LookupTable::from(&ProjectivePoint::conditional_select( &x_beta, &-x_beta, r2_sign, )); let digits1 = to_radix_16_half(&r1_c); let digits2 = to_radix_16_half(&r2_c); let mut acc = table1.select(digits1[32]) + table2.select(digits2[32]); for i in (0..32).rev() { for _j in 0..4 { acc = acc.double(); } acc += &table1.select(digits1[i]); acc += &table2.select(digits2[i]); } acc } impl Mul<Scalar> for ProjectivePoint { type Output = ProjectivePoint; fn mul(self, other: Scalar) -> ProjectivePoint { mul_windowed(&self, &other) } } impl Mul<&Scalar> for &ProjectivePoint { type Output = ProjectivePoint; fn mul(self, other: &Scalar) -> ProjectivePoint { mul_windowed(self, other) } } impl Mul<&Scalar> for ProjectivePoint { type Output = ProjectivePoint; fn mul(self, other: &Scalar) -> ProjectivePoint { mul_windowed(&self, other) } } impl MulAssign<Scalar> for ProjectivePoint { fn mul_assign(&mut self, rhs: Scalar) { *self = mul_windowed(self, &rhs); } } impl MulAssign<&Scalar> for ProjectivePoint { fn mul_assign(&mut self, rhs: &Scalar) { *self = mul_windowed(self, rhs); } }