Source-linked AI summary
Blockchain Censorship
Anton Wahrstätter, Jens Ernstberger, Aviv Yaish, Liyi Zhou, Kaihua Qin, Taro Tsuchiya, Sebastian Steinhorst, Davor Svetinovic, Nicolas Christin, Mikolaj Barczentewicz, Arthur Gervais
TL;DR
This paper asks how OFAC sanctions challenge censorship neutrality in permissionless blockchains and what censorship means for blockchain security. It formalizes and measures censorship across Bitcoin and Ethereum, finding substantial post-merge Ethereum censorship and delayed transaction confirmation. It also proves a threshold limitation on censorship resilience in PoS and PoW.
Problem
Permissionless blockchains are intended to rely on deterministic rules rather than third parties to decide transaction inclusion, but OFAC sanctions on a Bitcoin mixer and Ethereum application put that neutrality to the test.
Source: §1 · p. 1Method
The paper formalizes censorship across blockchain layers, quantifies OFAC-related censorship in Ethereum and Bitcoin, analyzes security effects, and proves an impossibility result for PoS and PoW.
Source: §1 · p. 2, §8 · p. 13Results
46% of post-merge Ethereum blocks were made by actors engaged in OFAC-related censorship; censorship delays transaction confirmation, and censorship resilience is impossible when censoring validators or miners exceed 50%.
Source: §8 · p. 13, §7 · p. 13Takeaways & Limitations
Censorship affects both blockchain neutrality and security by prolonging transaction confirmation, while a majority of censoring validators can prevent PoS censorship resilience.
Source: §8 · p. 13, §7 · p. 13Block-header timestamps provide only a rough estimate of confirmation time because miners and validators may not report precise block-generation times.
Source: §6.1 · p. 10
Abstract
from arXiv · showhide
Permissionless blockchains promise to be resilient against censorship by a single entity. This suggests that deterministic rules, and not third-party actors, are responsible for deciding if a transaction is appended to the blockchain or not. In 2022, the U.S. Office of Foreign Assets Control (OFAC) sanctioned a Bitcoin mixer and an Ethereum application, putting the neutrality of permissionless blockchains to the test. In this paper, we formalize quantify and analyze the security impact of blockchain censorship. We start by defining censorship, followed by a quantitative assessment of current censorship practices. We find that 46% of Ethereum blocks were made by censoring actors that intend to comply with OFAC sanctions, indicating the significant impact of OFAC sanctions on the neutrality of public blockchains. We further uncover that censorship not only impacts neutrality, but also security. We show how after Ethereum's move to Proof-of-Stake (PoS) and adoption of Proposer-Builder Separation (PBS) the inclusion of censored transactions was delayed by an average of 85%. Inclusion delays compromise a transaction's security by, e.g., strengthening a sandwich adversary. Finally we prove a fundamental limitation of PoS and Proof-of-Work (PoW) protocols against censorship resilience.
1 INTRODUCTION
The paper examines how OFAC sanctions test censorship neutrality in permissionless blockchains, then measures censorship and its security effects. It reports substantial post-merge Ethereum censorship, transaction delays, and a threshold beyond which PoS and PoW cannot ensure censorship resilience.
1 INTRODUCTION
- OFAC's 2022 sanctions targeted the Bitcoin mixer Blender.io and Ethereum's Tornado Cash, an autonomous, immutable application; the latter designation was unprecedented for smart-contract addresses.The paper reports that sanctions prevented Blender.io from continuing centralized services, while cautioning that the likely cause was removal of its website, not censorship as defined in the paper.Source: §1 · p. 1
- 46% of post-merge Ethereum blocks were made by actors engaged in transaction censorship due to OFAC sanctions over a two-month period.The paper also reports that after Ethereum's move to PoS and PBS, the average inclusion delay for Tornado Cash transactions rose from 15.8 ± 22.8 seconds in August 2022 to 29.3 ± 23.9 seconds in November 2022.Source: §1 · pp. 1–2
- OFAC sanctions were followed by an 84.3% decline in Tornado Cash contract interactions within two months and an 84.99% August spike in blocked users.Ethermine controlled 22% of Ethereum PoW hash rate and its refusal to include Tornado Cash interactions coincided with 200 fewer blocks containing those transactions per day (about 33.4%).Source: §1 · pp. 1–2
- Censorship increases the time transactions spend in the mempool, delaying both censored and non-censored transactions and increasing sandwich-attack risks.The paper identifies a greater-than-50% censoring share of validators or miners as an impossibility threshold for censorship resilience in PoS or PoW.Source: §1 · p. 2
- The paper formalizes blockchain censorship across system layers and measures censorship by smart contracts, builders, proposers, and relayers.It distinguishes consensus-layer obstruction from application-layer prevention of successful execution.Source: §1 · pp. 1–2
2 BACKGROUND
This background explains how permissionless chains, applications, and censorship pressures fit together. It describes the technical layers and OFAC-sanctions context relevant to the paper's analysis.
2.1 Permissionless Blockchains
- Permissionless blockchains use deterministic rules to govern transaction validity, while Bitcoin uses PoW and Ethereum moved to PoS in September 2022.Ethereum's smart contracts support DeFi, where transaction ordering can be exploited through sandwich, front-, and back-running attacks.Source: §2.1 · p. 2
- Proposer/Builder Separation divides block creation from block proposal, with relays intermediating between builders and proposers; Ethereum validators may opt out of PBS.PBS was introduced after Ethereum's PoS transition as a response to Miner/MEV problems and was supposed to enhance censorship resistance.Source: §2.1 · p. 2
- Privacy-enhancing services obscure asset flows; Tornado Cash uses fixed-denomination pools and zk-SNARKs, while Blender.io mixes deposits through a shared account.The fixed denomination is intended to make deposits harder to link to withdrawals.Source: §2.1 · pp. 2–3
2.2 Rationales for Censorship
- Censorship pressures may be external, such as government and legal demands, or internal, including ethical and economic motives; these can intertwine.The paper notes that legal uncertainty and operators' risk tolerance can shape censorship decisions.Source: §2.2 · p. 3
- Legal and political rationales for censorship include preventing transactions tied to criminal activity or economic-sanctions violations.The paper focuses on OFAC's 2022 designations of Blender.io and Tornado Cash; some Tornado Cash-listed addresses were smart contracts without administrative functionality.Source: §2.2 · p. 3
- The paper uses “OFAC-compliance” informally for actors' likely rationale, not as a claim that their conduct is legally required or sufficient for compliance.OFAC says the SDN list is not exhaustive, and whether validators, builders, or relays must censor remains debated.Source: §2.2 · p. 3
3 RELATED WORK AND CENSORSHIP EXAMPLES
Prior work studies censorship definitions, attacks, examples, and defenses. This paper's stated contribution is a quantitative overview of censorship across actors and its security implications.
3 RELATED WORK AND CENSORSHIP EXAMPLES
- Prior work defines blockchain censorship through transaction exclusion, announced intent to exclude, or refusal to attest to chains containing targeted transactions.The paper focuses on selective censorship within a network rather than censorship of entire networks.Source: §3 · pp. 3–4
- Censorship research covers consensus-layer feather-forking and network-layer vulnerabilities such as DNS seeding and IP hard-coding.Other studied attack vectors include denial of service, eclipse, routing, and prefix hijacking.Source: §3 · p. 4
- Examples span RPC endpoints, wallets, DeFi front ends, and mining pools, while prior defenses include evaluation frameworks, payment channels, mixers, congestion detection, and decentralization measurement.The paper notes that temporal execution delays may have significant financial implications for a censored entity.Source: §3 · p. 4
- The authors aim to quantify censorship imposed by different ecosystem actors and analyze its security implications.They state that, to their knowledge, prior work had not provided this combined quantitative overview.Source: §3 · p. 4
4 OVERVIEW OF CENSORSHIP
The paper models censorship across blockchain layers and defines transaction inclusion and censorship in terms of intentional obstruction. It distinguishes outright exclusion, delayed inclusion, and smart-contract blocking of execution.
4.1 System Model
- The overview extends a prior system model and treats the network, consensus, application, and auxiliary services as distinct system layers.The network layer covers validator communication and user transaction access; consensus appends ordered blocks, and smart contracts cause state transitions.Source: §4 · p. 4, §4.1 · p. 4
4.2 Notation & Terminology
- Transaction inclusion means a validator-received transaction appears in a ledger block after the protocol's finality delay; censorship resistance prevents censorship, while resilience tolerates individual acts of censorship.The notation models a blockchain maintained by validators running a state-machine replication protocol.Source: §4.2 · pp. 4–5
4.3 Definition of Censorship
- Strict censorship is intentional obstruction that keeps a transaction out of the blockchain, while weak censorship delays it past the next possible block.The paper excludes transaction reordering alone from censorship, although it notes that reordering can cause failure or lower user gain.Source: §4.3 · p. 5
- Smart contracts cannot directly prevent consensus-layer inclusion or finalization, but can block successful execution of an included transaction.A block list is an example of a contract preventing an account from successfully executing an interaction.Source: §4.3 · p. 5
5 CENSORSHIP QUANTIFICATION
The paper quantifies censorship across Ethereum’s consensus and application layers and Bitcoin, using data on sanctioned Tornado Cash and Blender.io interactions. It finds that sanctions coincided with sharp declines in activity, while Ethereum miners and post-PBS ecosystem participants differed in whether they included Tornado Cash transactions.
5.1 Data Collection
- The dataset covers Tornado Cash deposits and withdrawals from known Ethereum pool contracts and ecosystem blocks through November 15, 2022, but excludes other sanctioned addresses and contracts on Layer 2 or testnets.It contains 273,403 interactions in 236,868 distinct blocks; the authors state that most traffic from sanctioned entities is captured because users mainly use Ethereum mainnet ETH-denominated contracts.Source: §5.1 · pp. 5–6
5.2 Pre-PBS Consensus-Layer Censorship
- 16,347 interactions in July 2022 fell to 1,630 in October, but Tornado Cash activity never reached zero after sanctions.Weekly deposits and withdrawals had fallen from over 2,000 before sanctions to about 200 afterward; the authors also identify the website takedown, repository outage, and USDC freeze as possible explanations for the decline.Source: §5.2 · p. 6, Figure 1 · p. 7
- Ethermine’s uncensored blocks fell from 203 to about 21 per day after sanctions, a reduction of almost 90%, while other large miners declined by 50%–65%.No miner completely stopped including Tornado Cash transactions; Figure 2 shows uncensored blocks by the ten largest miners.Source: §5.2 · p. 7, Figure 2 · p. 7
5.3 Post-PBS Consensus-Layer Censorship
- After PBS, the four largest builders showed evidence of censoring Tornado Cash transactions, while only three of the most successful builders included them.Flashbots builders created about 22.2% of blocks in the analyzed period; the three non-censoring builders identified among the most successful belonged to BloXroute and Beaverbuild.Source: Figure 3 · p. 7, §5.3 · p. 8
- Bitcoin Suisse and Figment never included Tornado Cash transactions in the analyzed period, while Flashbots’ relay consistently did not forward blocks containing them.The two proposers accounted for almost 3% of proposed blocks; Flashbots relayed about 46% of all blocks proposed, while other relayers forwarded blocks containing Tornado Cash transactions.Source: §5.3 · p. 8
5.4 Application Layer Censorship
- 556 accounts were blocked by USDT from January through November 15, 2022, compared with 87 by USDC; after OFAC sanctions, Blender.io had no further interactions.The authors caution that Blender.io’s cessation likely followed removal of its website and was not censorship as defined in the paper.Source: §5.4 · p. 9, Figure 6 · p. 9, §5.5 · pp. 9–10
6 SECURITY IMPLICATIONS OF CENSORSHIP
Censorship can slow transaction confirmation and threaten security, while supplementary measurements test the reliability of the latency findings. The section also proves a threshold limit on censorship resilience in PoS.
6 SECURITY IMPLICATIONS OF CENSORSHIP
- Censorship slows confirmation by keeping transactions in the mempool longer, compromising security through risks including double-spending, sandwich attacks, and congestion.The section examines latency before and after Ethereum’s merge, distinguishing censored from non-censored transactions.Source: §6 · p. 10, §6.1 · p. 10
6.1 Transaction Confirmation Latency
- 29.3 ± 23.9 seconds versus 8.7 ± 8.3 seconds: November 2022 TC transactions waited longer than non-censored transactions; TC delay rose from 15.8 ± 22.8 seconds in August.Longer confirmation latency is associated with greater double-spending and sandwich-attack risk, among other security concerns.Source: §6.1 · p. 10
6.2 Supplementary measurement from an additional Ethereum node
- Both European and US nodes observed more than 95% of included transactions before inclusion, and both found TC transactions took longer to include than non-TC transactions.The TC/non-TC difference in observation time between nodes was statistically insignificant; that comparison did not control for gas fees.Source: §6.2 · p. 11
6.3 Impossibility of Censorship-Resilience
- More than 50% censoring validators makes censorship-resilience impossible for a PoS protocol in a synchronous network; the lower bound also applies to Nakamoto consensus.The proof relates censoring validators’ threat to liveness to that of adversarial validators.Source: §6.3 · pp. 12–13
7 DISCUSSION
The discussion describes how censorship incentives can propagate among proposers, relayers, and builders, and how execution-delay effects vary by application. It also states a PoS censorship-resilience threshold.
7 DISCUSSION
- Proposers can avoid participating in censorship by building locally, using only non-censoring relays, or setting MEV-Boost’s min-bid fallback.The fallback enables uncensored vanilla building when builder bids do not exceed a chosen threshold.Source: §7 · p. 13
- Relay censorship can pressure external builders and MEV searchers to produce censored blocks through relay access and financial incentives.Builders may be forced to comply when most proposers connect exclusively to one relay.Source: §7 · p. 13
- Execution-delay effects depend on the application: TC may lack deadline-related reverts, while delayed decentralized-exchange transactions can harm user experience.DeFi contracts may protect users from unfavorable execution prices when too much time passes.Source: §7 · p. 13
- More than 50% of PoS nodes adopting censorship practices prevents censorship-resilience; censoring majorities can withhold attestations from blocks that do not follow their rules.Only blocks adhering to the censoring validators’ rules are appended in this scenario.Source: §7 · p. 13
8 CONCLUSION
The paper formalizes and measures blockchain censorship, examines its security implications, and proves a censorship-resilience limit. It reports censorship in Ethereum and concludes that censorship can degrade security and user service quality.
8 CONCLUSION
- 46% of Ethereum blocks after the transition to PoS were made by actors intending to comply with OFAC sanctions.The paper also quantifies the effect of OFAC sanctions on Ethereum and Bitcoin.Source: §8 · p. 13
- Censorship prolongs transaction confirmation time, which the paper identifies as degrading blockchain security.The conclusion also characterizes censorship as degrading quality of service for users.Source: §8 · p. 13
- A blockchain cannot be censorship-resilient when more than 50% of validator nodes enforce censorship, according to the paper’s result.Source: §8 · p. 13