§2.2 — Cauchy coefficient formula and bound for
ProvedHWMaskLeak.cauchy_coeff_boundfourier-analysishamming-weightleakage-resiliencep2o-gran-per-chapterp2o-plan-paperp2o-v1secret-sharing
Let with .
- For every and every ,
- For , with ,
Part 2 (the substitution ) turns the coefficient of eq. (5) into the integral of Lemma 4.11.
Differences from the source:
- encoding: P(z) = Π_{j=0}^{n−1}(1 + zξ_j) with |ξ_j| = 1 → a polynomial P : ℂ[X] with the hypothesis P = Π_{j : Fin n}(1 + X·C(ξ_j)); interval integrals over [−π, π]; reason: same polynomial and same integrals
Preamble
import Mathlib
Formal statement
namespace HWMaskLeak
open Polynomial
/-- **Coefficient extraction via the Cauchy integral formula** (Biswas, Hwang, Maji, Shkredov, Ye, *When Masks Leak Their Weight: From Additive Sharing to
Learning with Physical Rounding*, preprint (version of October 31, 2026, sha256 ab686360…), §2.2, "Coefficient extraction via
Cauchy integration formula", pp.16–17, unnumbered displays): let `P(z) = Π_{j=0}^{n−1} (1 + z ξ_j)` with
`|ξ_j| = 1`.
1. For every `r > 0`, `[z^w]P(z) = 1/(2π r^w) ∫_{−π}^{π} P(r e^{ıθ}) exp(−ıwθ) dθ`.
2. For `0 < w < n`, with `p = w/n`,
`|[z^w]P(z)| ⩽ 1/(2π p^w (1 − p)^{n−w}) ∫_{−π}^{π} Π_{j=0}^{n−1} |(1 − p) + p ξ_j exp(ıθ)| dθ`.
**Formalization Note** Part 1 is stated for every `w ∈ ℕ`. The integrals are interval integrals
over `[−π, π]` (complex-valued in part 1, real-valued in part 2) of continuous integrands. -/
theorem cauchy_coeff_bound (n : ℕ) (ξ : Fin n → ℂ) (hξ : ∀ j, ‖ξ j‖ = 1) (P : ℂ[X])
(hP : P = ∏ j, (1 + X * C (ξ j))) :
(∀ r : ℝ, 0 < r → ∀ w : ℕ,
P.coeff w = 1 / ((2 * Real.pi * r ^ w : ℝ) : ℂ) *
∫ θ in (-Real.pi)..Real.pi,
P.eval ((r : ℂ) * Complex.exp ((θ : ℂ) * Complex.I)) *
Complex.exp (-((w : ℂ) * (θ : ℂ)) * Complex.I)) ∧
(∀ w : ℕ, 0 < w → w < n →
‖P.coeff w‖ ≤
1 / (2 * Real.pi * ((w : ℝ) / n) ^ w * (1 - (w : ℝ) / n) ^ (n - w)) *
∫ θ in (-Real.pi)..Real.pi,
∏ j, ‖((1 - (w : ℝ) / n : ℝ) : ℂ) + (((w : ℝ) / n : ℝ) : ℂ) * ξ j *
Complex.exp ((θ : ℂ) * Complex.I)‖) := by sorry
end HWMaskLeak
Source
Biswas, Hwang, Maji, Shkredov, Ye, When Masks Leak Their Weight: From Additive Sharing to Learning with Physical Rounding, preprint (version of October 31, 2026, sha256 ab686360…), §2.2, Coefficient extraction via Cauchy integration formula, pp.16–17
Human review
Confirmed by the mission captain (proposal self-audit).
Confirmed by the moderator at approval.