Sharp smoothing for the literal last successive minimum at the fixed negligible schedule
ProvedInfoLossLAR.WorstCase.RegevSketch.smoothing_le_lamN_logfixed-negligible-schedulegaussian-smoothingregevsuccessive-minima
For every positive dimension n and nonsingular integer input basis B, the actual Gaussian smoothing parameter of its lattice at the same fixed negligible schedule epsilon(n) used in the DGS sketch is at most (4/log 2) log(n+2) times the literal target-defined last successive minimum lambda_n(B). The constant is absolute and independent of the adversary, sample exponent and basis encoding length. The proof uses the infimum definition directly, without assuming that a short independent family is efficiently found. This is a smoothing bound, not an algorithmic DGS/SIVP/GapSVP reduction.
Preamble
import Definitions.Def_InfoLossLAR_IntegerBasis_Core import Definitions.Def_InfoLossLAR_RegevSketch_DGS import Theorems.Thm_InfoLossLAR_WorstCase_RegevSketch_smoothing_le_short_independent import Theorems.Thm_InfoLossLAR_WorstCase_RegevSketch_smoothingError_log_factor set_option autoImplicit false
Formal statement
namespace InfoLossLAR.WorstCase.RegevSketch
theorem smoothing_le_lamN_log {dimension : ℕ}
(basis : Matrix (Fin dimension) (Fin dimension) ℤ) (hdet : basis.det ≠ 0)
(hdimension : 0 < dimension) :
RegevLWE.GaussConv.smoothingParam (IntegerBasis.inputLattice basis hdet) (smoothingError dimension) ≤
(4/Real.log 2)*Real.log (dimension+2)*InfoLossLAR.WorstCase.lamN basis := by sorry
end InfoLossLAR.WorstCase.RegevSketch
Source
Complete existing checked proof GaussianPoissonSeries.smoothing_fixed_negligible_le_lamN_log and its original-basis/last-minimum helper proofs, extracted by official Lean syntax/info-tree oracle from Solutions/RegevBridgeFixedLogSmoothing.lean, RegevBridgeSharpSmoothingLastMinimum.lean, RegevBridgeIntegerBasis.lean, RegevBridgeIntegerDualCoordinates.lean, RegevBridgeSharpSmoothingParameters.lean, RegevBridgeNegligibleSmoothingError.lean and RegevBridgeLatticeInterfaces.lean. The already checked generic smoothing and logarithmic factor are reused through their formal APIs. Renaming is semantically checked by Lean definitional equality. No new mathematical argument.