discrete-gaussian

Discrete gaussian samples based on FACCT and DDLL13. Implemented with ~40 bits of accuracy (relative error <= 2^-40). Based on the analysis in WALT19 this implies an upper bound of 2^80 security in constructions.

Example

cargo run --example vartime --release

Sampling with target theta: 20
Approximated k = 23
Approximated theta: 19.534401406624458
1000 samples in: 16.90ms

Feature tracking

• variable time sampling in u32
  • in discrete-gaussian::FieldElement
  • in ring_math::PolynomialRingElement
• higher precision math
• cumulative distribution table
• constant time sampling in u32
  • in scalarff::FieldElement
  • in ring_math::PolynomialRingElement
• structure compatibility
  • ring_math::Matrix2D compat, sample into a matrix
  • ring_math::Vector compat, sample into a vector