Lemma 1, p. 29 — for fixed data collection ξ, the optimal protection is max{C_h(ξ), 0}
ProvedDigitalPrivacy.Strategy.lemma_1best-responsedata-privacyoptimizationp2o-batch-pfp2ap2o-gran-per-chapterp2o-plan-paperp2o-v1
Let , , and , and fix a data collection level . Let be the business's profit and
Then is the unique maximizer of over : for every ,
with equality only if .
This is the business's best protection response to a given collection level; Theorem 1's is this response evaluated at .
Formalization Note The paper allows ; the statement assumes , since contains and for , the profit increases in without bound being attained. Uniqueness is the reading of "the business sets". Footnote 29 (when ) is not part of the statement.
Preamble
import Mathlib import Definitions.Def_DigitalPrivacy_Strategy_Setting
Formal statement
namespace DigitalPrivacy.Strategy
/-- Lemma 1, p. 29: for a fixed data collection level `ξ ∈ [0, 1]`, the protection level
`max{C_h(ξ), 0}` is the unique maximizer of `C ↦ Π(ξ, C)` over `C ≥ 0`. -/
theorem lemma_1 (β ρ ψ P ξ : ℝ) (hB : Basic β ρ ψ P) (hξ : ξ ∈ Set.Icc (0 : ℝ) 1) :
0 ≤ max (Ch β ρ ψ P ξ) 0 ∧
∀ C : ℝ, 0 ≤ C →
profit β ρ ψ P ξ C ≤ profit β ρ ψ P ξ (max (Ch β ρ ψ P ξ) 0) ∧
(profit β ρ ψ P ξ C = profit β ρ ψ P ξ (max (Ch β ρ ψ P ξ) 0) →
C = max (Ch β ρ ψ P ξ) 0) := by sorry
end DigitalPrivacy.Strategy
Source
Fainmesser, Galeotti, Momot, Digital Privacy, author manuscript (This Draft: April 2022; SSRN 3459274), p. 29, Lemma 1; proof Online App. B
Human review
Confirmed by the mission captain (proposal self-audit).
Confirmed by the moderator at approval.