When a secret, such as a decryption key for archived data, is split among a committee, the main danger is often not a broken cipher. It is a group of members who are paid to pool their pieces. Blockchain systems handle this by making each member lock up a deposit that is confiscated ("slashed") if collusion is proven. They almost always charge every member the same deposit. That choice is safe when every member matters equally, as in ordinary "any $t$ of $n$" threshold sharing. Coding theorists, however, build secret sharing schemes from few-weight linear codes: error-correcting codes whose nonzero codewords take only two to five distinct sizes. These schemes deliberately produce hierarchical access structures, where some small groups can recover the secret and other larger groups cannot. In such schemes, members are not interchangeable. A uniform deposit then leaves some cheap coalitions underpriced.
This note proposes to make that observation exact. For a Massey secret sharing scheme, meaning the standard way of turning a linear code into a sharing scheme, the cheapest coalition an adversary can bribe is a small linear program over the code's codewords. When the code has few weights, that program can often be solved in closed form. Our central claim is that the code's structure decides whether symmetric slashing is optimal. If the lowest-weight codewords cover every participant evenly, uniform deposits are provably optimal. If they cluster, as we expect for codes built from posets that violate the classical minimality condition, there is a measurable bribery-cost gap that a code-aware deposit vector closes. Post-quantum security enters as the operating regime rather than the mechanism. Massey schemes are information-theoretic, so a quantum computer does not help an attacker against them. Once share transport is post-quantum, the economic price of collusion is the remaining attack surface. For data with decades-long confidentiality, it is also a long-lived one.
Core Claim
Fix a minimal linear code $C$ of length $n+1$ over $\mathbb{F}_q$. "Minimal" means no codeword's support strictly contains another's. When the scheme is built on the dual code, the minimal qualified coalitions are exactly the supports, with coordinate $0$ removed, of codewords of $C$ that are nonzero at coordinate $0$. Few-weight codes from defining sets are designed for this use. Linear Codes from Two Weakly Regular Plateaued Balanced Functions, Quaternary Codes and Their Binary Images and Optimal minimal linear codes from posets all propose Massey schemes on exactly these grounds. Because $C$ has small dimension $k$, the family $\mathcal{S}$ of minimal coalitions has at most $q^k$ members. Every member has size $w-1$ for one of the few weights $w$.
Suppose a rational member defects whenever the bribe exceeds their slashable deposit. This is the collateral logic of Decentralized Anonymous Micropayments and the rational-node model of Cryptoeconomic Security for Data Availability Committees. The designer then chooses deposits $d_i \ge 0$ with total stake $B$ to maximize the cheapest bribe:
$$t^* = \max_{d \ge 0,\ \sum_i d_i = B}\ \min_{S \in \mathcal{S}} \sum_{i \in S} d_i \;=\; B \cdot \min_{\lambda \in \Delta(\mathcal{S})}\ \max_{i}\ \Pr_{S\sim\lambda}[i \in S],$$
where $\lambda$ is a probability distribution over minimal coalitions. The second equality is linear-programming duality.
Uniform deposits achieve $B\,(w_{\min}-1)/n$. Here $w_{\min}$ is the smallest weight of a codeword that is nonzero at coordinate $0$. Since the expected coalition size under any $\lambda$ is at least $w_{\min}-1$, uniform deposits are optimal if and only if the minimum-size coalitions support a distribution with equal marginals on every participant, which is a fractional 1-design. This yields a concrete sufficient condition: the code's automorphism group acts transitively on coordinates $1,\dots,n$. That condition is plausible for many trace-based codes. In that case symmetric slashing, the design principle Cryptoeconomic Security for Data Availability Committees proves optimal for symmetric committees, carries over exactly.
When the condition fails, the ratio $t^*/t_{\text{unif}}$ measures how much security symmetric slashing gives away. We conjecture that two-level poset codes fail it in a structured way, with low-weight codewords concentrated on one level of the poset. These codes are minimal while violating the Ashikhmin–Barg condition $w_{\min}/w_{\max} > (p-1)/p$, which demands that codeword weights be close to one another. For such codes the optimal deposit vector should then be computable from the poset's generating function. Proving this for the two-level families is the main analytic task.
Research Program
The first strand is analytic. Each code family in the file comes with closed-form weight distributions. We will compute, for each one, the per-coordinate count of minimum-weight codewords nonzero at coordinate $0$. For the trace-based Weil-sum codes of Two-weight and three-weight linear codes based on Weil sums, we will check transitivity. That would certify symmetric slashing as optimal, giving a closed-form cheapest bribe of $B(w_{\min}-1)/n$. For two-level poset codes, the generating-function technique of Optimal minimal linear codes from posets and Zhu et al. should give the level-wise inclusion counts directly. We expect the optimal deposits to be constant on each level, which reduces the LP to two or three variables with a closed-form solution. The simplicial-complex codes of Quaternary Codes and Their Binary Images and Binary LCD Codes and Self-Orthogonal Codes via Simplicial Complexes sit in between. Inclusion there depends on which maximal faces contain a coordinate, so they test whether the theory generalizes.
The second strand extends the analysis to collusion against whistleblowers. Breaking Omertà proves that a whistleblowing mechanism with deposit $D$ is defeated once colluders post retaliation deposits above $(t-1)D$, where $t$ is the threshold. For hierarchical structures, $t$ is not a single number. We will show that the relevant quantity is the size of the cheapest deposit-weighted coalition. We will also show that non-uniform deposits change which coalition retaliation must target. This provides the per-coalition retaliation bound that the threshold analysis lacks.
The third strand covers the post-quantum regime. Post-Quantum Cryptography (PQC) Network Instrument documents harvest-now-decrypt-later exposure and slow server-side migration: 0.029% post-quantum adoption among 20.5 million OpenSSH connections, and over 83% of servers running software from 2019 or earlier. The matching economic risk is "harvest now, collude later." Members can keep their shares after their deposits unlock and sell them years afterward. We will model deposit lock duration as a design variable. We will connect it to the long-term-service result of Breaking Omertà, which says indefinitely held deposits can give unconditional economic security for transparent services. We deliberately do not propose new lattice primitives here. Post-quantum transport is taken as given, for example via the multisignature of A Dilithium-like Multisignature in Fully Split Ring and Quantum Random Oracle Model.
Baselines
The naive reference point is Shamir sharing, which is built on Reed–Solomon codes, with uniform deposits and threshold $t$ matched to the hierarchical scheme's smallest coalition. Its cheapest bribe is $tB/n$. It is the scheme a practitioner would deploy by default. It answers whether hierarchy is worth having at all, given the same total stake $B$ and the same committee size $n$.
The strongest existing method is symmetric no-reward slashing from Cryptoeconomic Security for Data Availability Committees, applied to the same few-weight code. That paper's simulation reports that reducing query success from 100% to 99.9% requires about \$3.9 million in bribes per query, with 300,000 nodes at 32 ETH stake each. We will reuse those stake parameters so our cost-per-unit-stake numbers can be read against that scale. The \$3.9 million figure is for data withholding, not secret reconstruction. We found no reported result on Lacuna for bribery cost of code-based or Massey secret sharing schemes, so our comparison is on the same codes under the same budget. We beat this baseline if $t^*/t_{\text{unif}}$ is materially above one, say at least 1.2, for some family that is actually used. We also beat it in a different sense if our analysis certifies $t^*/t_{\text{unif}} = 1$ in closed form where the baseline could only assume it.
The oracle is the brute-force LP over the explicitly enumerated family $\mathcal{S}$. Enumeration is feasible because $q^k$ is small, for example $3^6 = 729$ codewords for the $[112,6,72]$ ternary code. The oracle is ground truth for checking our closed forms. It also bounds how much any heuristic can gain.
The heuristic baseline sets each deposit proportional to how often that participant appears in minimum-weight coalitions. It is what a careful engineer would try without the duality argument.
The key ablation tests whether the design property drives the result. We will compare pairs of codes, or randomly relabeled access structures, that share the same weight distribution but differ in automorphism structure. If the gap tracks transitivity rather than weight spread, the design condition is the operative ingredient.
Why not just…?
Why not just use Shamir sharing, where uniform slashing is trivially optimal? For pure threshold needs, Shamir wins and we will say so; its automorphism group is the full symmetric group, so it is a degenerate case of our theorem. The proposal matters only where applications need hierarchical access, the structures that Optimal minimal linear codes from posets and Quaternary Codes and Their Binary Images build. The Shamir baseline comparison shows what hierarchy costs in bribery resistance per unit of stake.
Why not just solve the deposit LP numerically and skip the coding theory? This is a real threat. For codes of small dimension, the brute-force LP is cheap and exact, and we use it as our oracle. The coding theory pays off in three places. It gives certificates and closed forms that hold for whole infinite families rather than one instance at a time. It gives parameter guidance before a code is chosen. And it remains usable at dimensions where enumerating $q^k$ codewords becomes infeasible. If the closed forms add nothing beyond the LP at deployable sizes, the honest conclusion is that the contribution reduces to the duality characterization, and the experiments will show which case holds.
Why not just adopt the threshold whistleblowing analysis of Breaking Omertà unchanged? That analysis is stated in terms of a single threshold $t$. For hierarchical structures it either overstates security, by using the typical coalition size, or understates it, by using the smallest. Our second strand computes the exact bound for these structures, and the simulation compares it against the threshold-substituted bound.
Proposed Experiments
We will instantiate four code families: the $[112,6,72]$ ternary Weil-sum code, which Two-weight and three-weight linear codes based on Weil sums reports one short of the Griesmer bound of 73; binary two-level poset codes across a sweep of level sizes; Gray images of simplicial-complex quaternary codes; and Reed–Solomon/Shamir as a reference. The weakly regular plateaued construction is excluded because its authors report no known instance.
For each family we will enumerate $\mathcal{S}$, solve the oracle LP, and report three quantities: $t^*/B$, $t_{\text{unif}}/B$, and their ratio. We will also run an agent-based bribery game in which an adversary buys the cheapest coalition at Tas–Boneh stake parameters. A second version of the game adds smart-contract retaliation against whistleblowers.
The decisive checks are two. First, transitive trace codes should show a ratio of exactly one, confirming the design theorem. Second, poset codes should show a ratio above one that grows as the gap between $w_{\min}$ and $w_{\max}$ widens, and that is matched by the level-wise closed form. If poset codes also show a ratio of one, the asymmetry concern raised by violating Ashikhmin–Barg is economically vacuous. That would itself be worth reporting, because it would mean symmetric slashing is safe across every code family in the literature we examined.
References
- Linear Codes from Two Weakly Regular Plateaued Balanced Functions
- Quaternary Codes and Their Binary Images
- Optimal minimal linear codes from posets
- Decentralized Anonymous Micropayments
- Cryptoeconomic Security for Data Availability Committees
- Two-weight and three-weight linear codes based on Weil sums
- New Quaternary Codes Derived from Posets of the Disjoint Union of Two Chains
- Binary LCD Codes and Self-Orthogonal Codes via Simplicial Complexes
- Breaking Omertà: On Threshold Cryptography, Smart Collusion, and Whistleblowing
- Post-Quantum Cryptography (PQC) Network Instrument: Measuring PQC Adoption Rates and Identifying Migration Pathways
- A Dilithium-like Multisignature in Fully Split Ring and Quantum Random Oracle Model