💻 computer science

Position Paper: Denial-of-Service Against Multi-Round Transaction Simulation

This paper introduces evasive, low-cost denial-of-service attacks that exploit multi-round transaction simulation and inter-transaction dependencies to disrupt Ethereum bundling services like Flashbots, significantly reducing builder revenue and slowing block production, while also proposing mitigation strategies.

Original authors: Yuzhe Tang, Yibo Wang, Wanning Ding, Jiaqi Chen, Taesoo Kim

Published 2026-04-24

📖 4 min read☕ Coffee break read

CC BY 4.0

Original authors: Yuzhe Tang, Yibo Wang, Wanning Ding, Jiaqi Chen, Taesoo Kim

Original paper licensed under CC BY 4.0 (http://creativecommons.org/licenses/by/4.0/). ✨ This is an AI-generated explanation of the paper below. It is not written or endorsed by the authors. For technical accuracy, refer to the original paper. Read full disclaimer

Imagine the Ethereum blockchain as a massive, high-stakes auction house where people are constantly trying to buy and sell digital items.

In this auction house, there are special workers called Block Builders. Their job is to gather all the pending bids (transactions) from the crowd, organize them into a neat package (a block), and hand that package to the auctioneer (the validator) to be officially recorded.

To make the most money, these builders use a clever system called Flashbots Bundles. Instead of just grabbing bids randomly, they run a multi-round rehearsal.

Round 1: They try out a few different orders of bids to see which one makes the most profit.
Round 2: They tweak the order again.
Round 3: They try again.
Final Round: Once they find the perfect order, they lock it in and send it to the auctioneer.

During these rehearsal rounds, the builders are essentially "pretending" to execute the transactions to see what happens. If a bid fails in the rehearsal, they throw it out before the final show.

The Problem: The "Ghost" Attack

The researchers in this paper discovered a way to hack this rehearsal system. They found a way to make the builders waste all their time and energy during the final, real round, without the builders even realizing they were being attacked until it was too late.

Here is how they did it, using a simple analogy:

The Analogy: The "Trap Door" in a Play

Imagine the Block Builder is a stage director rehearsing a play with a cast of actors (the transactions).

The Old Way (Previous Attacks):
Before this paper, hackers tried to break the play by having an actor scream loudly or run around wildly during the rehearsal.
- Why it failed: The director (the builder) is smart. If an actor screams during the rehearsal, the director just says, "Okay, that's a bad take," and throws that actor out of the rehearsal. The play continues smoothly, and the hacker gets kicked out.
The New Way (This Paper's Attack):
The researchers invented a "Trojan Horse" strategy. They created a sequence of actors (transactions) that work together like a Rube Goldberg machine.
- Actors 1 to N-1: These actors enter the stage and do nothing but set up a hidden trap. They whisper to each other, "Okay, the stage is ready." The director watches them, sees they are behaving perfectly, and lets them stay.
- The Trap: These actors have set up a secret condition. They are waiting for a specific signal that only happens in the Final Round.
- The Final Actor: When the director finally says, "Okay, this is the real show, let's go!" (the final round), the secret signal is triggered. Suddenly, the first N-1 actors all start screaming, running, and causing chaos at the exact same time.

Why This is Dangerous

Evasive: Because the actors behaved perfectly during the rehearsals (Rounds 1 through N-1), the director never kicked them out. They were allowed into the final, real show.
Inter-Dependent: The chaos only happens because the actors were talking to each other and sharing a secret state. Previous hackers tried to use "lone wolf" actors who acted alone, which didn't work. These new hackers used a "team" approach.
The Result: The director is now stuck dealing with a massive, chaotic mess in the middle of the real show. The show slows down to a crawl, the director loses money, and the other actors (normal users) can't get their lines in.

The Takeaway

The paper shows that by using a team of coordinated transactions that hide their malicious intent until the very last moment, attackers can crash the system that builds blocks.

Cost: It's very cheap for the attacker to set up.
Risk: It's "risk-free" because the attackers don't lose any money if the attack fails; they just look like normal users until the trap springs.
Impact: It slows down the entire blockchain ecosystem and hurts the people running the auction house.

The authors conclude that while we can try to build better "security guards" (mitigations) to catch these teams, the fundamental design of running multiple rehearsal rounds makes it very hard to stop this specific type of trick. It's a new kind of vulnerability that no one had seen before because everyone was looking for lone wolves, not coordinated teams.

Based on the provided position paper, here is a detailed technical summary of the research on Denial-of-Service (DoS) attacks against Multi-Round Transaction Simulation in Ethereum.

1. Problem Statement

The paper addresses a critical security gap in Ethereum's block-building ecosystem, specifically targeting transaction-bundling services (e.g., Flashbots Bundles) used by MEV (Maximal Extractable Value) searchers.

Context: Block builders assemble candidate blocks from unconfirmed transactions. To maximize revenue, modern builders (especially bundling services) utilize a multi-round execution model. In this model, transactions are simulated across multiple rounds with distinct orderings to filter out reverting transactions and optimize bundle composition.
The Gap: Existing DoS attacks against block builders (such as ConditionalExhaust, GhostTX, and DoERS) rely on speculative execution or resource exhaustion. However, these attacks are largely ineffective against multi-round builders.
- Reason for Failure: Existing attacks use passive checks to detect if execution is speculative. They trigger resource-exhaustion payloads prematurely during early simulation rounds (pre-rounds). Because these early rounds often run in parallel or isolated threads, the induced slowdown does not affect the final block production or normal transactions. The payload is "wasted" before the critical final round where transactions are executed serially.

2. Methodology: Evasive, Risk-Free DoS Attacks

The authors propose a novel, proactive attack design that exploits inter-transaction dependencies via shared smart contract state to bypass the multi-round simulation defenses.

Core Mechanism: State-Guarded Payloads

Instead of using independent transactions, the attack employs a sequence of transactions that interact with a shared smart contract state.

The Setup: The attacker deploys a contract containing a resource-exhaustive payload (e.g., heavy computation). This payload is hidden behind a conditional branch guarded by $n-1$ distinct state variables.
The Sequence: The attacker issues a sequence of $>n$ transactions:
- Transactions $1$ to $n-1$ : These transactions execute normally in the early simulation rounds. They update the specific state variables required to unlock the next condition. Because they appear normal and do not trigger the heavy payload, they pass all simulation checks in pre-rounds.
- Transaction $n$ (and subsequent): Once the state is fully prepared (after $n-1$ rounds), the final transaction triggers the resource-exhaustive payload.
The Trigger: The payload is only executed in the final round of the builder's simulation. In this final round, the builder executes transactions in serial order to construct the actual block.
The Impact: Because the heavy computation occurs in the final serial round, it directly delays the processing of all subsequent transactions in that block, causing significant slowdowns and reducing the builder's throughput.

Key Attack Characteristics

Evasiveness: The attack remains invisible during the multi-round filtering process because the "heavy" part of the transaction only activates when the specific state conditions are met, which only happens in the final round.
Risk-Free: The attack does not require the attacker to lose funds (e.g., no need to pay for failed transactions or gas in a way that results in financial loss).
Low Cost: It leverages the builder's own optimization features (atomic block inclusion) to amplify the impact without requiring massive capital investment.

3. Key Contributions

Identification of a New Attack Vector: The paper identifies that inter-transaction dependencies (transactions sharing state across a sequence) are a previously unexplored vector for DoS attacks in the context of multi-round builders. Prior works relied on sequences of independent, isolated transactions.
Novel Attack Design: The authors developed the first successful DoS attacks specifically tailored for multi-round bundling services, demonstrating that passive checks are insufficient for defense.
Mitigation Strategies: The paper explores the defense landscape and proposes mitigation strategies. It highlights a fundamental challenge: securing any finite-round builder against DoS attacks is inherently difficult because the builder must eventually execute transactions serially to produce a block, creating a window for exploitation.

4. Experimental Results

The authors tested their attacks against Flashbots' bundling service (the only known open-source bundling service).

High Success Rate: The attacks successfully triggered the resource exhaustion in the final round.
Revenue Reduction: The attacks substantially reduced the revenue of block builders by preventing them from including profitable transactions due to time constraints.
Block Production Slowdown: The attacks caused significant delays in block production times, degrading the overall performance of the network.

5. Significance

This research fundamentally shifts the understanding of DoS threats in Ethereum's MEV ecosystem.

Paradigm Shift: It moves the attack vector from "flooding" or "censorship compliance" to state-dependent sequencing.
Systemic Risk: It demonstrates that the very mechanisms designed to optimize block building (multi-round simulation and atomic inclusion) can be weaponized to disrupt the ecosystem.
Future Implications: The findings suggest that current defenses are insufficient for multi-round architectures. It necessitates a re-evaluation of how builders handle transaction ordering and state simulation, potentially requiring new consensus-level or protocol-level changes to prevent state-based DoS attacks.

In summary, the paper proves that by chaining transactions through shared state, an attacker can "hide" a denial-of-service payload until the exact moment it causes maximum damage to the block builder, effectively neutralizing current multi-round defense mechanisms.