# Accountable Transaction Inclusion Lists: Enhancing Ethereum's Censorship Resistance

Crypto.world

> In Ethereum, the problem is usually not whether a transaction gets included, but how long it is delayed, and time-critical applications such as auctions, stop-loss orders, and commit-and-reveal…

**arXiv**: 2607.22040

**Published**: 2026-07-24

**Authors**: Patrick Spiesberger, Hannes Hartenstein

## AI summary

## Overview

**Research area:** Blockchain systems security — censorship resistance and transaction inclusion mechanisms for Ethereum, specifically inclusion-list designs and their game-theoretic accountability.

**Technical level:** Intermediate to Advanced. The paper assumes familiarity with Ethereum's slot/proposer/attester structure, Proposer–Builder Separation (PBS), MEV, and basic mechanism-design reasoning, though the core argument (unaccountable committees are cheap to bribe; accountable ones are not) is stated in plain terms.

**Scope:** The paper formalizes accountability for transaction exclusion decisions, proposes the Fair Forward Inclusion Lists (FairFIL) mechanism, and empirically prices the cost of censorship against FIL, FOCIL, MCP, and AUCIL using Ethereum mainnet measurements from September to December 2025.

## What This Paper Is About

In Ethereum, the problem is usually not whether a transaction gets included, but how long it is delayed, and time-critical applications such as auctions, stop-loss orders, and commit-and-reveal schemes suffer when a block builder suppresses a transaction across consecutive slots. Existing proposals (notably FOCIL, scheduled for deployment in late 2026) split inclusion control across a committee of κ validators, but no committee member is punished for omitting a transaction, so an adversary can bribe the whole committee cheaply. The paper asks whether making exclusion decisions publicly disclosed, verifiable, and attributable to a specific party substantially raises the cost of censorship, and answers with a concrete mechanism.

## Key Contributions

1. **A formalization of accountability for censorship resistance.** The authors define inclusion-list-based censorship resistance mechanisms (IL-CR), distinguish "constructing" mechanisms (committee picks list contents freely, hence non-accountable) from "verifying" mechanisms (a protocol-anchored construction policy is checked by a committee), and prove that accountable enforcement raises censorship costs by roughly one order of magnitude relative to prior approaches.
2. **The FairFIL protocol.** A Π₁ (one-slot forward) accountable mechanism in which every block assembler must publish all transactions it censors, a validator committee checks the disclosure's completeness and validity against its mempool view, and the next assembler forfeits the full block reward ℛ(B_{s+1}) for any omission.
3. **A proof that compliance is rational.** Within the Byzantine, Altruistic, Rational (BAR) behavior model (with altruistic participants assumed absent), the authors show FairFIL enforces block invalidation under sustained censorship and that compliance is rational for all participants.
4. **An empirical evaluation on Ethereum mainnet.** The implementation quantifies per-slot censorship costs, compares them with FIL, FOCIL, MCP, and AUCIL, checks the mempool-consistency assumption, and assesses practical deployability while preserving most builder MEV-extraction freedom.

## Main Findings

- **Censorship is measurable in practice:** in September 2025, an average of 8.3 transactions per block were censored despite being eligible for inclusion under the intended tip-based reference ordering.
- **Builder concentration motivates the problem:** two builders assembled more than 80% of all blocks in 2024 and 2025, and nearly all proposers rely on them.
- **Una accountable committees are cheap to bribe:** for FOCIL, the per-slot bribery cost across the entire κ = 16 committee is below 2 € for a median-tip transaction in late 2025; AUCIL, the most effective of the fee-based mechanisms, still allows censorship at approximately 3.30 € per slot expected cost.
- **Cumulative censorship costs differ sharply (Table 2, median-tip transaction, Sep.–Dec. 2025 measurements):** baseline Ethereum ≈0.10 € / ≈0.20 € / ≈1 € for 1 / 2 / 10 slots; FIL ≈0.11 € / ≈0.22 € / ≈1.10 €; FOCIL ≤1.54 € / ≤3.08 € / ≤15.40 €; MCP ≤1.60 € / ≤3.20 € / ≤16 €; AUCIL ≤3.30 € / ≤6.60 € / ≤33 €.
- **FairFIL under the "attack suppression" path:** ≈0.10 € for one slot, ≈50.34 € for two slots, ≈402 € for ten slots.
- **FairFIL under the "invalidation" path:** ≈0.10 € for one slot, ≈28.30 € for two slots, ≈141 € for ten slots (and the paper notes invalidating a block can be significantly more expensive due to MEV extraction).
- **Extending censorship beyond one slot costs a full block reward:** under FairFIL, a censoring assembler must forfeit ℛ(B_s), so sustained censorship is priced by the block reward rather than by a transaction tip.
- **MEV freedom is largely preserved:** FairFIL constrains the assembler only by requiring that any transaction may be censored for one slot but must be included thereafter.
- **Mempool consistency appears sufficient:** initial measurements suggest the consistency FairFIL requires in practice is met (Section 5.2), supporting the identical-mempool-view assumption used in the analysis.
- **Baseline affordability:** a budget of 0.10 € per slot suffices to censor the median transaction under plain Ethereum in the observed period, and that cost is predominantly driven by the fixed on-chain bribe overhead c_bribe.
- **Committee-bribing is out of reach:** the slot committee comprises approximately 30,000 validators in 2025, so bribing a majority to accept an invalid or reject a valid block is argued to exceed the block reward.

## Methodology in Plain English

The authors start by defining what counts as censorship: a transaction is censored from a block if it was propagated before the slot started (operationalized as at least 3 s before slot start) and would have fit under a deterministic "Fair Ordering" — mempool transactions sorted by descending per-gas tip with the transaction hash as tie-breaker, admitted greedily while block validity holds. They then model all validators and block assemblers as rational: any strictly positive revenue gain induces deviation, while ties are resolved in favor of compliance. The adversary is not profit-driven, can message every participant, pays bribes on-chain through a bribing contract (incurring a fixed overhead c_bribe), and is budget-capped so it cannot stop block publication entirely.

Against that model, they derive the per-slot cost of censoring a transaction for each existing mechanism and compare it with FairFIL. FairFIL's design has two moving parts: the assembler of block B_s publishes FairFIL_s, built by scanning mempool transactions first seen within the last 603 s (fifty slots of 600 s plus a 3 s propagation margin), sorting them by Fair Ordering, discarding transactions that are already in B_s or that would not be includable, then re-executing the remaining censored transactions on the post-B_s state and keeping only those that stay valid. Validators of slot s vote down the block if any transaction from FairFIL_{s−1} is missing, if FairFIL_s deviates from the prescribed ordering or is not executable, or if the disclosure is incomplete relative to their own mempool view. Finally, the authors implement FairFIL and run it against Ethereum mainnet data from September to December 2025 to price censorship and to test whether real mempools are consistent enough for the verification to work.

## Why This Matters

**Impact on research:** the paper reframes censorship resistance from a question of committee size to a question of accountability, showing that a mechanism can be made expensive to subvert without throttling MEV extraction. It supplies a formal vocabulary (constructing vs. verifying mechanisms, (ℳ, σ)-censorship resistance, four accountability properties) that later work can build on, and it moves the debate beyond the one-of-κ-honest assumption toward rational-participant analysis.

**Real-world applications:**

- On-chain auctions, where a bid submitted just before the closing slot can be suppressed past the deadline.
- Decentralized finance, where a stop-loss transaction triggered at a price can be delayed while the price moves further against the user.
- Commit-and-reveal protocols, where a censored reveal is indistinguishable from an intentional non-reveal and the participant can be penalized despite acting in time.
- General time-critical transactions that need inclusion within a handful of slots rather than eventual inclusion.

**Industry relevance:** FOCIL is scheduled for Ethereum deployment in late 2026, so the paper's cost comparison lands directly on an upcoming protocol decision. It also engages with the reality that over 90% of validators voluntarily adopted MEV-Boost despite censorship and centralization risks, which is why the authors deliberately keep FairFIL MEV-friendly: mechanisms that cut into builder revenue risk being ignored by rational operators.

## Future Directions

- **Deployment and client engineering:** turning FairFIL into a production-ready protocol change, including how the 603 s lookback window and disclosure list interact with real block-building pipelines and private order flow.
- **Mempool consistency at scale:** the paper offers only "initial measurements" that the required consistency holds; broader measurement across node implementations and network conditions is an open empirical question.
- **Alternative penalty structures:** the analysis prices omission as forfeiture of the full block reward; whether partial penalties, escalating penalties, or other attribution schemes change the rational-compliance result is unexplored here.
- **Adversary models beyond the current budget cap:** the authors explicitly exclude arbitrarily wealthy adversaries, self-censorship, and censorship at the peer-to-peer layer, leaving those regimes open.

## Target Audience

Ethereum protocol researchers and client developers, mechanism-design and blockchain-security academics, and technically literate practitioners (staking operators, builders, and application developers running time-critical protocols) who want to understand how inclusion lists could be made accountable before FOCIL-style mechanisms ship.

## Authors’ abstract

In Ethereum, transaction inclusion is rarely in question; what matters is the delay until inclusion. Currently, block builders could exercise censorship across consecutive blocks, threatening time-critical applications, such as on-chain auctions. To mitigate this risk, existing proposals such as FOCIL, scheduled for deployment in late 2026, assign a committee to list transactions for mandatory inclusion. However, no committee member is held accountable for the actual…

[The full abstract on arXiv](https://arxiv.org/abs/2607.22040)

[Read the original paper](https://arxiv.org/abs/2607.22040)

---

Canonical: https://crypto.world/paper/accountable-transaction-inclusion-lists-enhancing-ethereum-s-censorship-2607.22040
