Skip to content
Crypto.world

Research

Concave is the New Linear: The Impossibility of Anti-Plutocratic DAO Governance

DAOs govern on-chain protocols by letting token holders vote, and the standard rule — one token, one vote — concentrates power in a few large wallets. Communities have proposed "anti-plutocratic"…

Concave is the New Linear: The Impossibility of Anti-Plutocratic DAO Governance
arXiv
2605.18990
Published
2026-05-18
Authors
Austin Bennett, Preston Vander Vos, Duc V. Le, Mira Belenkiy

AI summary

Overview

Research area: Algorithmic game theory and blockchain security — specifically the mechanism design of Decentralized Autonomous Organization (DAO) governance.

Technical level: Intermediate. The headline result is stated in plain economic terms (splitting tokens across wallets restores linear voting power), but the formal treatment uses Jensen's inequality, constrained optimization, and asymptotic analysis.

Scope: The paper proves that no wallet-balance-based voting rule can resist a Sybil attacker on a permissionless blockchain, and quantifies the practical cost of such attacks against five major DAOs.

What This Paper Is About

DAOs govern on-chain protocols by letting token holders vote, and the standard rule — one token, one vote — concentrates power in a few large wallets. Communities have proposed "anti-plutocratic" fixes like Quadratic Voting, which gives each wallet the square root of its token balance in votes, so that splitting holdings across wallets ought to be discouraged. This paper asks whether any such wallet-level rule survives an attacker who simply creates many wallets, and answers that it cannot: concavity only buys a constant-factor discount, never a sublinear one.

Key Contributions

  1. A unified, costed model of wallet-based voting. The authors formalize an arbitrary voting rule f alongside a cost scheme C = (m, v, p, s) covering a minimum voting balance m, per-wallet voting cost v, a one-shot setup cost p, and a per-wallet splitting cost s. They define "plutocratic" as the asymptotic slope of the attacker's Sybil-adjusted voting power, V*(a), giving one definition that applies uniformly to linear, concave, and other rules.

  2. A universal impossibility result. They prove that every nontrivial wallet-based voting rule is plutocratic under unrestricted splitting — if any wallet size yields nonzero voting power, an attacker holding a tokens achieves V*(a) = Ω(a) — and that this holds under any linear cost scheme.

  3. A tight asymptotic slope for concave rules. For finite, positive, increasing, concave rules, they sharpen the lower bound to an exact constant: V*(W) ≤ κ(a − p) with κ = sup_{x ≥ m} f(x)/(x + v + s), asymptotically tight as the attacker's budget grows.

  4. An empirical instantiation on real DAOs. They replay the ten most recent finalized proposals of five DAOs (ENS, Compound, and Uniswap on Ethereum, plus Arbitrum and ZKsync) with per-proposal gas prices, under linear, quadratic, logarithmic, and power (β = 0.25) voting, and measure realized Sybil amplification and attack costs.

Main Findings

  • Every nontrivial wallet rule is plutocratic. The Sybil-Induced Plutocracy theorem shows that if some wallet size m has f(m) > 0, then for all a ≥ 2m there is a constant α > 0 with V*(a) ≥ α · a. A cheap concrete bound: splitting into equal m-sized chunks gives V*(a) ≥ ⌊a/m⌋ f(m).

  • Costs do not rescue the mechanism. The Robustness of Plutocracy to Costs theorem shows that adding a fixed setup cost p, per-wallet splitting cost s, per-wallet voting cost v, and minimum balance m still yields V*_C(a) = Ω(a). Costs change the constant, not the growth order.

  • Even splits are the optimal attack. By Jensen's inequality, for a concave f and a fixed wallet count n, total power ∑ f(w_i) is maximized by the equal allocation w_i = a/n. The attacker's problem therefore reduces to choosing n.

  • The per-wallet optimum is a closed form. With the minimum balance slack, the optimal balance is w* = (a − p)/n − v − s, giving aggregate power V*(W) = n · f((a − p)/n − v − s), maximized over feasible n.

  • Concavity collapses to linearity asymptotically. For concave rules, V*(W) ≤ κ(a − p) with κ = sup_{x ≥ m} f(x)/(x + v + s), and lim_{a→∞} V*(W)/(a − p) = κ when the supremum is attained. Geometrically, κ is the slope of the steepest line through the origin lying above the attacker-return curve. A single honest wallet, by contrast, earns only f(a) — sublinear whenever f curves.

  • Closed forms exist for the proposed rules. For quadratic voting the worst-case Sybil form is V*(W) = (a − p) / (2√(v + s)), versus V(W) = √a honestly; the paper's Table 1 collects the linear, quadratic, and power-voting expressions, with power voting (β ∈ (0,1)) scaling as (a − p)(β^β)(1 − β)^(1 − β) / (v + s)^(1 − β). The logarithmic entry of that table is not present in the available text.

  • Real amplification factors are large. Under Quadratic Voting, measured Sybil amplification ranges from 1,172× to 4,039× across the five DAOs; under a steeper power rule with β = 0.25, amplification exceeds 229,000×.

  • Attack cost is orders of magnitude below value at stake. On Uniswap, a Sybil-optimal attacker captures a $300M vote for roughly $75K in token and gas costs, while a single-wallet attacker with the same budget reaches only ≈ 1/259 of the voting power needed to win.

  • The problem is empirically grounded. The introduction notes that only about 5% of total token supply participates in governance votes on average, and cites a systematization finding 28 real-world attacks across four blockchains, roughly evenly split across bribing, token control, human-computer interaction, and code/protocol vulnerability.

Methodology in Plain English

The authors build a mathematical model of an attacker with a governance tokens who can split them into any number of wallets. Each wallet costs something to create and to vote from — gas for an ERC-20 transfer, gas for a castVote transaction — and some DAOs require a minimum balance before a wallet may vote at all. The attacker's objective is the total votes they can extract, and the authors ask how that total scales as the attacker's budget grows.

They then prove three things in sequence. First, a crude argument: splitting into many equally sized wallets always yields voting power proportional to holdings, so no rule escapes. Second, for concave rules, the best split is an even one (by Jensen's inequality), and the best wallet size follows from the binding budget constraint. Third, optimizing over the number of wallets produces a single constant κ that fully characterizes the attacker's asymptotic return per token.

To test the theory, they take real on-chain data — the ten most recent finalized proposals from each of five DAOs, together with per-proposal gas prices on each chain — and compute what an optimizing attacker would actually achieve and pay under each voting rule. Deliberately, the model includes no utility function for the attacker: the authors argue that real attackers measure success outside the system, so worst-case voting power, not token value, is the right thing to bound.

Why This Matters

Impact on research. The result closes an entire design space rather than critiquing one mechanism. Quadratic Voting was proposed precisely because linear voting is plutocratic; the paper shows the proposed cure restores the disease asymptotically. It also reframes the question: defenses must limit wallet count, not per-wallet behavior, which redirects work toward identity-based Sybil resistance.

Real-world applications:

  • Treasury security for DeFi protocols. ENS, Compound, Uniswap, Arbitrum, and ZKsync all govern significant assets; the Uniswap figure — a $300M vote for roughly $75K — is a direct statement of exposure.
  • Governance token design. Teams choosing between linear, quadratic, and power voting now have a quantified worst case rather than a heuristic.
  • Attack-cost budgeting for delegates and foundations. The framework lets a protocol estimate what an adversary would pay to reach quorum under its specific gas costs and thresholds.
  • Post-mortem analysis. The cited incidents — the May 2023 Tornado Cash proposal that replaced the governance contract, and the November 2023 Indexed Finance attacks that forced a defensive proposal permanently locking the treasury — fit the "token control" attack family the paper formalizes.

Industry relevance. The paper's conclusion is prescriptive for practitioners: per-wallet economic friction only reduces the slope κ, never its order. Anti-plutocratic governance on a permissionless chain requires an identity layer limiting wallet count, and concave voting should be deployed only as one component of such a composite mechanism, never as the primary line of defense.

Future Directions

  • Designing the identity layer. The paper argues that Sybil resistance must bound wallet count, but does not specify how — Proof-of-Personhood, attestation schemes, and their compositions are surveyed as candidate defenses rather than analyzed as theorems.
  • Composite mechanisms. If concave voting is only one component of a larger design, what combinations of identity gating plus concave weighting actually achieve sublinear attacker power, and at what usability cost?
  • Non-concave and non-standard rules. The tight κ result covers finite, positive, increasing, concave rules; the authors explicitly set aside convex, disjoint, and other non-standard rules as requiring bespoke attacks, leaving their optimal attack structure open.
  • Non-linear token transfer fees. The cost model assumes fixed per-wallet costs, justified by the fact that all standard ERC-20 transfers are linear; tokens that charge a percentage of transferred value are noted as existing but are not analyzed, and the authors state they know of no DAO governance token charging non-fixed voting fees.

Target Audience

Protocol designers and governance-token engineers choosing a voting mechanism; DAO delegates and treasury risk managers who need to quantify capture cost; and mechanism-design researchers working on Sybil resistance and collusion-resistant voting. The paper is also useful to security auditors assessing governance attack surfaces, and to readers who want a rigorous counterargument to the intuition that concave voting is inherently more democratic than one-token-one-vote.

Authors’ abstract

Decentralized Autonomous Organizations (DAOs) run protocol governance by letting token holders vote on proposals. The dominant rule, voting power proportional to wallet balance, concentrates control among a small number of large holders, fueling the token-control governance attacks that have already compromised real protocols. To counter this concentration, the community has turned to anti-plutocratic voting mechanisms such as Quadratic Voting (QV), which assign sublinear…

The full abstract on arXiv
Read the original paper