Class BLS12_381Pairing

java.lang.Object
org.bouncycastle.crypto.bls.BLS12_381Pairing

public class BLS12_381Pairing extends Object
Optimal ate pairing on BLS12-381: a bilinear, non-degenerate map e: G1 x G2 -> GT where GT is the order-r subgroup of Fp^12 ^*.

This implementation favours obvious correctness over performance:

  • G2 is lifted into E(Fp^12) via the D-twist isomorphism (x', y') -> (x'/w^2, y'/w^3).
  • The Miller loop runs the textbook affine doubling / addition formulas on E(Fp^12) and computes the line evaluations as full (non-sparse) Fp^12 elements.
  • The final exponentiation is performed as a single Fp^12 modPow with exponent (p^12 - 1) / r, sidestepping the Frobenius-coefficient infrastructure that the spec-recommended easy/hard-part split would need.
Sparse line evaluation, Frobenius-based final exponentiation, and a Jacobian-coord G2 in pairing context are all natural follow-on optimisations that preserve the public surface here.
  • Field Summary

    Fields
    Modifier and Type
    Field
    Description
    static final BigInteger
    Hard-part exponent (p^4 - p^2 + 1) / r (~1269 bits), applied after the Frobenius-based easy part.
  • Method Summary

    Modifier and Type
    Method
    Description
    The hard part of the final exponentiation, exposed for cross-package layered testing (the test classes live in org.bouncycastle.crypto.hash2curve.test and need direct access to the easy/hard split for KAT comparison against reference outputs).
    multiPair(ECPoint[] g1Points, BLS12_381G2Point[] g2Points)
    Multi-pairing: compute the product e(P_0, Q_0) * e(P_1, Q_1) * ... * e(P_{n-1}, Q_{n-1}) with a single shared Miller loop and a single final exponentiation.
    Compute the optimal ate pairing e(P, Q) for P on BLS12-381 G1 and Q on BLS12-381 G2.

    Methods inherited from class Object

    clone, equals, finalize, getClass, hashCode, notify, notifyAll, toString, wait, wait, wait
  • Field Details

    • HARD_EXPONENT

      public static final BigInteger HARD_EXPONENT
      Hard-part exponent (p^4 - p^2 + 1) / r (~1269 bits), applied after the Frobenius-based easy part. The full final exponent (p^12 - 1) / r (~4317 bits) factors as (p^6 - 1) * (p^2 + 1) * (p^4 - p^2 + 1) / r; the easy part computes f^((p^6 - 1)(p^2 + 1)) essentially for free using conjugation and one Frobenius² application, leaving only this shorter exponent for Fp12Element.modPow(BigInteger).
  • Method Details

    • pair

      public static Fp12Element pair(ECPoint g1, BLS12_381G2Point g2)
      Compute the optimal ate pairing e(P, Q) for P on BLS12-381 G1 and Q on BLS12-381 G2.
      Parameters:
      g1 - a point on the BLS12-381 G1 curve. Must be in the prime-order subgroup; this method does not subgroup-check.
      g2 - a point on the BLS12-381 G2 curve. Must be in the prime-order subgroup; this method does not subgroup-check.
      Returns:
      e(P, Q) as an Fp^12 element in the order-r subgroup of Fp^12 ^*. Returns 1 if either input is the point at infinity.
    • multiPair

      public static Fp12Element multiPair(ECPoint[] g1Points, BLS12_381G2Point[] g2Points)
      Multi-pairing: compute the product e(P_0, Q_0) * e(P_1, Q_1) * ... * e(P_{n-1}, Q_{n-1}) with a single shared Miller loop and a single final exponentiation. This cuts a 2-pairing verification (e.g. BLS signature verify) to roughly the cost of one pair() call, since the dominant final exponentiation is performed only once.

      Pairs whose G1 or G2 component is the point at infinity are skipped (their pairing value is 1, identity in GT). If all pairs are skipped, the result is Fp12Element.ONE.

      Parameters:
      g1Points - G1 inputs.
      g2Points - G2 inputs; must be the same length as g1Points.
      Returns:
      the product of pairings as an element of GT.
      Throws:
      IllegalArgumentException - if the arrays differ in length.
    • hardPart

      public static Fp12Element hardPart(Fp12Element f)
      The hard part of the final exponentiation, exposed for cross-package layered testing (the test classes live in org.bouncycastle.crypto.hash2curve.test and need direct access to the easy/hard split for KAT comparison against reference outputs). Not part of the intended public API of this class — production callers should use pair(ECPoint, BLS12_381G2Point) / multiPair(ECPoint[], BLS12_381G2Point[]).