Imagine you're mixing cocktails. One bartender free-pouring vodka gives you a decent random pour — but a trained palate can detect the pattern. Now have three independent bartenders each free-pour a shot and add them together. The combined pour is so irregular that even a taster with superhuman senses (read: a quantum computer) can't distinguish it from a perfectly uniform random splash. That's the core mechanism here: summing independent random permutations over a finite group to create a function that looks indistinguishable from a truly random function, and proving exactly how many 'tastes' (queries) an adversary needs before they can tell the difference. The committed claim: a single Fourier-analytic framework simultaneously derives tight classical and quantum indistinguishability bounds for the sum of k ≥ 2 independent random permutations over any finite abelian group of order N. This is not the first paper to study the sum-of-permutations construction — Patarin's H-coefficient technique and prior combinatorial methods have been the workhorses — but it is the first to unify classical and quantum security analysis under one representation-theoretic roof, and the first to give concrete quantum security bounds for the construction and its variants. The ladder here matters. Classically, the bound Ok(q/N^{k−1/2}) for all q < N matches or improves upon the best prior results by Patarin and subsequent refinements, and the sub-birthday refinement to Ok(q²/Nᵏ) is new in this generality. For quantum security, prior work (Hosoyamada–Iwata, Bhaumik–Nandi–Jha) gave asymptotic bounds for k = 2 via Zhandry's recording technique. This paper gives concrete finite-N bounds via Fourier interpolation up to q ≤ 4N/15, and for k ≥ 3 obtains Ok(q³/Nᵏ) across the full range, which is genuinely new. The matching one-query Fourier attack achieving Θ(N⁻²) and the N/2-query parity attack with advantage 1/2 confirm that these bounds hit the right query thresholds — you can't improve the exponents by much. The architecture is pure mathematics: no computation, no simulation, no hardware dependency. The paper lives in the Fourier-analytic family of cryptographic proof techniques. The key structural choice is representing both the construction's probability density and the distinguisher's acceptance function in the Fourier domain over the group algebra, then exploiting the fact that classical queries restrict the distinguisher to pointwise evaluations (bounded Fourier support) while quantum queries allow superposition access (wider but still structured Fourier support). A simulation lemma then bridges the quantum query model to the Fourier bounds. This is the Fourier-analytic counterpart to Patarin's H-coefficient technique and Zhandry's compressed oracle / recording method — a genuinely different proof technology applied to the same objects. Integrity is strong in the way pure mathematics demands: results are proved, not simulated. The bounds come with explicit constants and matching attacks. The one-query Fourier attack and the N/2-query parity attack serve as lower bounds, confirming tightness. There is no dataset, no benchmark, no code — which is appropriate for this kind of work. The potential weakness is that 'concrete bounds' still involve implicit constants in the Ok notation for k ≥ 3, but for k = 2 the bounds are fully explicit. The paper also extends to two important variants: truncated / linearly-postprocessed sums (covering practical constructions like nonce-misuse-resistant encryption) and Dinur's LXoP construction, giving the first quantum security proofs for both. The milestone question for this line of work is not about hardware scaling but about proof coverage. The immediate next target is extending these Fourier bounds to non-abelian groups (e.g., the symmetric group SN itself), which would cover a broader class of practical block cipher constructions. The paper explicitly works over finite abelian groups, and the Fourier analysis relies on the character theory of abelian groups. Extending to non-abelian settings would require representation-theoretic machinery of a different order. The other concrete milestone: proving quantum security of the full Dinur construction for arbitrary output widths, not just fixed ones. The experiment the authors did not run — and here 'experiment' means 'proof' — is the non-abelian extension. The honest read is (a): it's genuinely harder, not a matter of compute budget but of mathematical difficulty. The representation theory of non-abelian groups does not decompose as cleanly, and the Fourier interpolation step would require new ideas. They also did not attempt to remove the q ≤ (N−1)/2 constraint on quantum queries, which likely requires a fundamentally different simulation argument. These are real open problems, not withheld results.
