# PDoS: A Profitable Denial-of-Service Attack against Proof-of-Work Blockchain Liveness

Crypto.world

> Prior attacks split into two families: withholding/infiltration attacks that earn the attacker money but do not try to halt the chain, and liveness attacks such as BDoS that can deter rational miners…

**arXiv**: 2609.12450

**Published**: 2026-09-11

**Authors**: Junjie Hu, Tianzhu Han, Na Ruan

## AI summary

PDoS: A Profitable Denial-of-Service Attack against Proof-of-Work Blockchain Liveness

## Overview

**Research area:** Blockchain security and consensus-layer incentive analysis — specifically attacks on Proof-of-Work (PoW) liveness that exploit mining-pool reward mechanics and miner economic rationality.

**Technical level:** Advanced. The paper relies on continuous-time Markov chains (CTMCs), Markov reward processes, steady-state analysis, and game-theoretic equilibrium arguments.

**Scope:** The paper introduces PDoS, a hybrid attack that pairs header-only "signal deterrence" with parasitic share-reward extraction from an infiltrated victim pool, and analytically shows that disrupting PoW liveness can be cost-neutral or profitable for the attacker.

## What This Paper Is About

Prior attacks split into two families: withholding/infiltration attacks that earn the attacker money but do not try to halt the chain, and liveness attacks such as BDoS that can deter rational miners from mining but force the attacker to absorb continuous losses. This made liveness denial-of-service economically unsustainable in high-value networks, particularly after the April 2024 reward halving. The paper asks whether a single adversary can degrade chain liveness *and* have the victim pool's share-reward mechanism pay for the attack, and answers yes with PDoS.

## Key Contributions

1. **A new attack class.** PDoS is presented as the first denial-of-service attack against PoW liveness that is economically self-sustaining and can generate a positive accounting profit in real-world settings, whereas prior mining attacks achieve either profitability or liveness disruption but not both (Table 1).
2. **A unified analytical framework over three target responses.** The authors formalize system evolution for target strategy S ∈ {Mine, SPV, Stop} and derive closed-form steady-state and utility expressions covering attacker–victim–target interaction.
3. **A dynamic, MEV-aware incentive model.** The model combines the coinbase reward K, a transaction-fee accumulation rate λ_t, and bursty "whale" rewards arriving at rate λ_w with lognormal magnitude. Revenue is decoupled into an MEV inflation coefficient e^S, effective block-generation rate v_M^S, share-extraction rate s_M^S, and active-time ratio θ_M^S, yielding an analytical break-even characterization and a proof that the target subgame converges to a unique strict Nash equilibrium in which all target miners stop mining.
4. **Trace-driven validation spanning the 2024 Bitcoin halving.** Relative to BDoS, PDoS is reported to lower the attacker hash-power threshold and reduce net attack cost, making "infinite attack endurance" easier to reach, expanding the feasible attack region via a lower operating break-even bound and a higher validity upper bound, and quantifying PPLNS and delayed-payout defenses against a no-hardening baseline.

## Main Findings

- **Liveness can be self-funded.** The attacker splits hash power between private mining (α_pri) and infiltration of the victim pool (α_inf), submitting shares (PPoWs) while withholding full blocks. The victim pool's proportional share payout subsidizes the attack, so the "attack subsidization" effect lowers the hash power needed to push rational miners offline.
- **Threshold dominance (Theorem 4.2).** Under identical network and economic parameters, the critical attacker hash power for PDoS is never larger than for BDoS: α_A*|PDoS ≤ α_A*|BDoS. A smaller share of hash power suffices to flip the target's best response from Mine to Stop.
- **Net cost dominance (Theorem 4.4).** The minimum net-cost rate satisfies C_A*|PDoS(α) ≤ C_A*|BDoS(α). When C_A* ≤ 0 the attacker has crossed break-even and the attack becomes victim-funded and self-sustaining.
- **High fees or high MEV can weaken security.** This is framed as counterintuitive: in high-fee or high-MEV environments, a higher block value increases the attacker's parasitic revenue, pushing the attack across the break-even point into a self-sustaining or profitable regime rather than strengthening the chain.
- **Defensive retreat amplifies the attack.** The reward-inflation coefficient e^S = 1 + M(Δα_act^S / (1 − Δα_act^S)) ≥ 1, where M ∈ [0,1) is the MEV share of block value, ordered as e^Stop > e^SPV = e^Mine. Because shutdown creates the largest active-hash-power vacuum, blocks mined during deterrence are worth more, and the deterring-state probability ordering π_det^Mine < π_det^SPV < π_det^Stop (Lemma 3.1) means shutdown prolongs the very states the attacker wants.
- **SPV mining is strictly dominated.** The paper proves U_T^SPV < U_T^Mine under all parameter settings: SPV output during deterrence is invalid because the parent body is withheld (the "SPV trap"). Iterated elimination of strictly dominated strategies leaves S ∈ {Mine, Stop}.
- **Victim-power hijacking in races.** Race win probabilities are p_A^3 = α + γ(η + δ) for private-vs-victim, p_A^4 = α + γ(β + η + δ) for private-vs-external, and p_A^5 = α + β + γ(η + δ) for infiltration-vs-external, where γ ∈ [0,1] is the attacker's rushing ability (γ = 0.5 is symmetric). In the infiltration race the victim pool is effectively forced to support the attacker's branch.
- **Reported comparison status.** Per Table 1, withholding attacks are marked profitable but not liveness-disrupting, generic DoS attacks and BDoS are liveness-disrupting but not profitable, and PDoS is marked both.
- **What is not reported in the provided text.** The excerpt is truncated mid-way through Definition 4.5 (Attack Endurance), so the quantitative evaluation results — measured break-even hash-power values, endurance figures, and the PPLNS / delayed-payout defense numbers — are not reported here. The paper also notes that PDoS can theoretically offset explicit costs under BDoS given race-winning block rewards and MEV, but only at an unrealistically high hash-power threshold and extreme market conditions.

## Methodology in Plain English

The authors model the network as four groups whose hash power sums to one: the attacker (α), a victim pool (β) that always validates full blocks and pays shares proportionally, a continuum of rational target miners (η), and other honest miners (δ).

The attack is a small state machine with six states: state 0 (normal chain growth), states 1 and 2 (the attacker has a lead from private mining or from inside the victim pool and broadcasts only the block *header*, withholding the body), and states 3, 4, 5 (races triggered when someone else finds a block — private-vs-victim, private-vs-external, infiltration-vs-external). They solve the continuous-time Markov chain for the long-run fraction of time spent in each state, then attach rewards to those states with a Markov reward process.

Two behavioural rules drive the attack. First, the attacker reallocates hash power by state: an infiltration fraction r₁ with private mining 1 − r₁ in the initial state, r₂ with zero private mining while deterring, all power to the private branch in states 3 and 4, and all power into the victim pool in state 5. Second, if targets keep mining normally, the attacker releases the withheld body to force a race; if targets try SPV mining on the header, the attacker discards the body so their work is worthless.

Economics are expressed per unit of hash power: normalized utility U_M^S = c[ω_b(v_M^S + s_M^S) − θ_M^S], where c is operating cost per unit hash power per unit time, ω_b = λR_b/c is the baseline profitability factor, v is effective block generation, s is parasitic share extraction (zero for targets), and θ is the fraction of time online. Hardware purchase costs are explicitly excluded as sunk. Attack success is then defined through two thresholds — the critical hash power α_A* at which the target's shutdown utility gap ΔU_T^Stop turns negative, and the minimum net-cost rate C_A* — and the theorems compare both against BDoS. Validation is trace-driven, covering the April 2024 Bitcoin reward halving, with PPLNS and delayed-payout defenses benchmarked against a no-hardening baseline (numerical outcomes are outside the provided excerpt).

## Why This Matters

**Research impact.** The paper reframes MEV and fee growth as a security liability rather than a security subsidy: because the attacker's revenue scales with block value, richer blocks can move an attack from loss-making to self-sustaining. It also unifies two previously separate literatures — pool-payout exploitation and liveness deterrence — into one Markov-reward framework, and shifts the analytical lens from "attack at any cost" to accounting profit including race-winning rewards and MEV.

**Real-world applications.**

- **Mining-pool payout design:** motivates auditing proportional share-reward schemes against share-withholding infiltration, and motivates the PPLNS and delayed-payout mitigations the paper evaluates.
- **Protocol and client hardening:** header-only relay is already normal in Bitcoin, so the deterrence signal is cheap to emit; the SPV trap argues for tighter coupling between header acceptance and body availability in lightweight validation.
- **Miner risk management:** gives rational miners a quantitative rule for when to shut down versus keep mining under elevated orphaning risk.
- **Security budgeting and monitoring:** provides operators with threshold metrics (α_A*, net cost rate, endurance under budget W) to judge whether a network's fee/MEV profile makes liveness attacks economically viable.

**Industry relevance.** Mining-pool industrialization and rising fee and MEV shares make block rewards heterogeneous, so payout-rule exploitation is a practical rather than theoretical concern. The authors responsibly disclosed the findings to the Bitcoin Core Security Team in April 2026 and withheld detailed disclosure records until the vulnerability handling process concludes.

## Future Directions

- **Optimal attacker policy.** The optimization of the infiltration ratios r₁ and r₂ is deferred to Appendix F, leaving the optimal dynamic allocation schedule as an explicitly open piece of the analysis.
- **Stricter pool policies.** The authors flag that attack success depends on stale-share acceptance and propagation delay (captured through γ), and defer discussion of stricter pool policies to Section 7 — an invitation to test whether delayed payouts or stale-share rejection close the subsidy channel.
- **Endurance under budget constraints.** Definition 4.5 introduces attack endurance T_A* for an initial budget W, but the truncated text cuts off before the endurance results; relating budget, endurance, and MEV conditions quantitatively remains the natural next step.
- **Richer adversary and market models.** The model assumes no difficulty retargeting within the attack horizon, ignores ordinary propagation delay, and treats MEV via a single ratio M plus bursty lognormal whale rewards; extending to retargeting, real propagation, and multi-pool victim structures would test how far the profitability boundary moves.

## Target Audience

Researchers in blockchain security and consensus incentives; mining-pool operators and protocol developers who design or audit share-reward and relay rules; and quantitatively trained analysts (game theory, stochastic processes, or mechanism design) assessing whether fee- and MEV-rich PoW networks remain economically secure. Readers need comfort with Markov chains and utility modeling; newcomers will find the qualitative story — "the victim pool pays for the attack" — accessible, but the theorems and closed-form results require an advanced background.

## Authors’ abstract

The security and liveness of Proof-of-Work (PoW) blockchains fundamentally depend on the economic rationality of miners. Existing incentive-driven denial-of-service attacks, such as BDoS, can deter rational miners from participating, but require the attacker to continuously absorb substantial economic losses and are therefore difficult to sustain in high-value networks. Meanwhile, prior infiltration-based withholding attacks are designed to extract revenue rather than to…

[The full abstract on arXiv](https://arxiv.org/abs/2609.12450)

[Read the original paper](https://arxiv.org/abs/2609.12450)

---

Canonical: https://crypto.world/paper/pdos-a-profitable-denial-of-service-attack-against-proof-of-work-2609.12450
