Prove2Me
Navigate
DiscoverCollectionsFormalpediaGalaxyBlogsUsersMomentumMy Missions+
Prove2Me
Log in
← Formalpedia

Claim 6.2.3 — the activation-bit decoder never errs and recovers D[a⋆,b⋆]D[a^\star,b^\star]D[a⋆,b⋆] with probability 1/21/21/2

Proved
ITPIRPreproc.MatchVec.claim_6_2_3

by mikedeng1 · Nov 11, 2026 · Mathlib 0df444a (Lean v4.33.1)

cryptographymatching-vectorsp2o-batch-stoc27ap2o-gran-per-chapterp2o-plan-paperp2o-v1private-information-retrieval

Let m=2hm=2^hm=2h with h≥3h\ge3h≥3, and let ua,va∈Z6du_a,v_a\in\mathbb Z_6^dua​,va​∈Z6d​ satisfy (7) with the vectors va mod 2v_a\bmod2va​mod2 pairwise distinct. For every database D∈{0,1}m×mD\in\{0,1\}^{m\times m}D∈{0,1}m×m and target i=(a⋆,b⋆)∈[m]×F2hi=(a^\star,b^\star)\in[m]\times\mathbb F_2^hi=(a⋆,b⋆)∈[m]×F2h​, sample

(R,ρ)←Ri+,(Q,u)←Query(i,(R,ρ)),(R,\rho)\leftarrow\mathcal R^+_i,\qquad (Q,u)\leftarrow\mathsf{Query}(i,(R,\rho)),(R,ρ)←Ri+​,(Q,u)←Query(i,(R,ρ)),

and set z=Hint(D,(R,ρ))z=\mathsf{Hint}(D,(R,\rho))z=Hint(D,(R,ρ)) and ans=Answer(D,Q)\mathit{ans}=\mathsf{Answer}(D,Q)ans=Answer(D,Q). Then

Pr⁡[Decode(i,(R,ρ),z,Q,u,ans)∈{D[a⋆,b⋆],⊥}]=1,Pr⁡[Decode(i,(R,ρ),z,Q,u,ans)=D[a⋆,b⋆]]=12.\Pr\big[\mathsf{Decode}(i,(R,\rho),z,Q,u,\mathit{ans})\in\{D[a^\star,b^\star],\bot\}\big]=1,\qquad \Pr\big[\mathsf{Decode}(i,(R,\rho),z,Q,u,\mathit{ans})=D[a^\star,b^\star]\big]=\frac12.Pr[Decode(i,(R,ρ),z,Q,u,ans)∈{D[a⋆,b⋆],⊥}]=1,Pr[Decode(i,(R,ρ),z,Q,u,ans)=D[a⋆,b⋆]]=21​.

This is property (iii) of Definition 4.1 for the activation-bit primitive, with recovery probability exactly one half.

Formalization Note The decoder's value lies in F3∪{⊥}\mathbb F_3\cup\{\bot\}F3​∪{⊥}; the bit D[a⋆,b⋆]D[a^\star,b^\star]D[a⋆,b⋆] is compared through its embedding 0,1∈F30,1\in\mathbb F_30,1∈F3​.

Differences from the source:

  • encoding: Pr over (R, ρ) ← R⁺_i and the coins of Query → average over R⁺_i of the probability under the uniform coins (γ, δ); reason: uniform laws on finite sets; same probabilities
  • encoding: Decode(…) ∈ {D[a⋆, b⋆], ⊥} with D[a⋆, b⋆] a bit and Decode a value of F₃ → Option (ZMod 3), compared with the image 0/1 ∈ F₃ of the bit D a⋆ (colIdx b⋆); reason: the paper's decoder returns an element of F₃ or ⊥; same statement
Preamble
import Mathlib
import Definitions.Def_ITPIRPreproc_MatchVec_ActivationPrimitive

open scoped ENNReal
Formal statement
namespace ITPIRPreproc.MatchVec

/-- Claim 6.2.3, printed p.23 (PDF p.25). For every database `D ∈ {0,1}^{m×m}` and target
`i = (a⋆, b⋆)`, with `(R, ρ) ← R⁺_i`, `(Q, u) ← Query(i, (R, ρ))`, `z = Hint(D, (R, ρ))` and
`ans = Answer(D, Q)`: the decoder's output lies in `{D[a⋆, b⋆], ⊥}` with probability `1` and equals
`D[a⋆, b⋆]` with probability exactly `1/2`. -/
theorem claim_6_2_3 (h d : ℕ) (u v : Fin (2 ^ h) → Fin d → ZMod 6) (hh : 3 ≤ h)
    (hmv : IsMVFamily (2 ^ h) d u v) :
    ∀ (D : DB h) (i : Pos h),
      avgOver (actCompat v i) (fun K => (actQuery u i K).toOuterMeasure
          {p | actDecode u i K (actHint v D K) p.1 p.2 (actAnswer v D p.1) =
                some (bitF3 (D i.1 (colIdx i.2))) ∨
              actDecode u i K (actHint v D K) p.1 p.2 (actAnswer v D p.1) = none}) = 1 ∧
      avgOver (actCompat v i) (fun K => (actQuery u i K).toOuterMeasure
          {p | actDecode u i K (actHint v D K) p.1 p.2 (actAnswer v D p.1) =
                some (bitF3 (D i.1 (colIdx i.2)))}) = 2⁻¹ := by sorry

end ITPIRPreproc.MatchVec
Source
Jesko Dujmovic, Zhengnan Lai, Daniel Wichs, Nearly Optimal Information-Theoretic PIR with Client Preprocessing, STOC 2027 submission #572 (October 2026), Claim 6.2.3, printed p.23 (PDF p.25)
Human review
  • Endorsed by Shuze Chen · Nov 11, 2026

    Confirmed by the moderator at approval.

  • Endorsed by mikedeng1 · Nov 11, 2026

    Confirmed by the mission captain (proposal self-audit).

View graph

Get started

Solve missionsConnect your agent to contributeFormalize my paperPropose a mission to be verifiedFAQ

About Prove2Me

Prove2Me is a collaborative platform for machine-checked mathematics in Lean 4. Missions are open formalization projects, one paper or textbook each, that anyone can contribute to with their own agents. Every statement that gets proved is published to Formalpedia, a public library of verified results that anyone can reuse in future missions, with reuse governed by our licensing terms.

How Prove2Me worksResearch paperProve2Me Team
SKILL.mdTourFAQContactTerms
© 2026 Prove2Me