# Incentive Attacks in BTC: Short-Term Revenue Changes and Long-Term Efficiencies

Crypto.world

> Bitcoin's DAA keeps block production roughly constant per epoch, and this adjustment is what allows withholding-style strategies to eventually raise an attacker's share of rewards.

**arXiv**: 2511.11538

**Published**: 2025-11-14

**Authors**: Mustafa Doger, Sennur Ulukus

## AI summary

## Overview

**Research area:** Blockchain security and mechanism design — specifically incentive attacks on Bitcoin's Proof-of-Work consensus and its Difficulty Adjustment Algorithm (DAA).

**Technical level:** Advanced. The paper uses Markov-chain style attack-cycle analysis, Catalan-number combinatorics, closed-form probability expressions, and numerical optimization.

**Scope:** A rigorous short-term and long-term comparison of the revenue and efficiency effects of selfish mining, intermittent mining, coin hopping between BTC and BCH, and block withholding (PAW-Type-B) attacks on adversarial, pool, and non-pool honest miners.

## What This Paper Is About

Bitcoin's DAA keeps block production roughly constant per epoch, and this adjustment is what allows withholding-style strategies to eventually raise an attacker's share of rewards. The paper asks two questions the literature has treated unevenly: what happens to *per-hashpower revenue* in the short term (before and around the difficulty adjustment), and what happens to *long-run efficiency* (revenue per cost per time) for both the attacker and the honest miners. It answers these for a family of selfish mining strategies, for coin hopping between BTC and Bitcoin Cash (BCH), and for a block withholding attack the authors call PAW-Type-B.

## Key Contributions

1. **A closed-form short-term analysis for general selfish mining.** The authors use *L-selfish mining* (originally introduced as *L-stubborn mining*) as a tractable substitute for the ε-optimal selfish mining strategies obtained via Markov Decision Process models. They note that L*-selfish mining revenue ratios are within 10^-2 of the ε-optimal values, which lets them study short-term revenue changes and profit lags for all parameters without running MDPs or Monte Carlo simulations.
2. **A new result on who benefits from coin hopping.** For a hopper switching between BTC and BCH at difficulty-adjustment boundaries, the paper shows that loyal honest BTC miners gain the *same* positive short-term revenue change per unit of computational power as the hopper, and that this increase is proportional to the adversarial fraction of hashpower.
3. **PAW-Type-B, a withholding model without fork simplifications.** The authors replace the simplifying assumption used in prior Fork After Withholding (FAW) and Power Adjusting Withholding (PAW) work — that each fork race resolves in favor of the adversarial full PoW with a fixed probability — with an explicit fork race using the same network-influence parameter γ as the selfish mining model. They derive closed-form expected adversarial and pool revenue.
4. **A new long-term efficiency metric and a strategy boundary.** They define efficiency as revenue/cost per hashpower per time for attacker and honest miners, and use it to establish a boundary between selfish mining and coin hopping, identifying which strategy is optimal for each parameter. They also study alternative adversarial objective functions beyond revenue ratio, including ones that guarantee the attacker gets the most out of the attack among all miners, and one based on revenue change just before the difficulty adjustment.

## Main Findings

- **Intermittent mining adds little in the short term:** The short-term benefits of intermittent mining strategies are described as negligible compared to the original selfish mining attack (Eyal and Sirer, corresponding to L = 2 in this framework).
- **Selfish mining wins in the long term:** Over the long run, selfish mining provides better efficiency than the intermittent alternatives studied.
- **Coin hopping is not purely adversarial in the short term:** Under a BTC–BCH coin hopping attack, loyal honest BTC miners see the same positive short-term revenue change per unit of computational power as the hopper.
- **A clear long-term boundary exists:** The paper establishes a new boundary between selfish mining and coin hopping, identifying the optimal efficient strategy for each parameter.
- **Block withholding helps outsiders most:** For block withholding strategies, honest miners *outside* the pool profit from the attack, usually even more than the attacker, in both the short term and the long term.
- **PAW has no profit lag:** A Power Adjusting Withholding attacker does not necessarily observe a profit lag in the short term — even without a difficulty adjustment, a PAW attacker makes profits. Since the selfish mining profit lag has long been treated as a main reason the attack is not seen in practice, the authors argue that barrier does not apply to PAW, and that relatively small pools are at immediate threat.
- **Doubt cast on an existing claim:** The authors state their long-term framework casts doubt on the claim that orphan reporting guarantees honest mining is the most profitable strategy in PoW systems, because that claim ignores an open market where miners can leave if they make losses.

*Note: the provided excerpt ends partway through Section II-C. Specific numerical values from the later analysis sections (III–VI) are not reported in the available content.*

## Methodology in Plain English

The authors avoid simulation wherever possible and instead build attack models whose quantities can be written down as exact formulas.

- **Attack cycles as the unit of analysis.** Each strategy is broken into repeating cycles — a period starting when the miners share a common block and ending when a new common chain tip is established. Within a cycle they track how many blocks the adversary mines, how many enter the eventual longest chain, and how many are wasted.
- **A parameterized selfish mining family.** The parameter L controls how long the adversary keeps mining privately against the honest chain. Setting L = 1 gives honest mining, L = 2 gives the classic Eyal–Sirer selfish mining attack, and L = ∞ gives equal-fork stubborn mining. Larger L values are optimized to approximate the best possible selfish mining policy, using Catalan-number expressions for the cycle probabilities.
- **Key measurable quantities.** They compute the block redundancy ratio (total blocks generated per canonical block) and the revenue ratio (fraction of canonical blocks the adversary earns), from which the number of epochs and the timing of difficulty adjustments follow.
- **Long-term framing with an ideal open market.** To compare strategies fairly on efficiency, they assume a stabilized open market in which honest miners who remain under the attack have a sustainable revenue–cost relationship. This lets them put all strategies on one consistent footing instead of comparing conflicting claims under different models.
- **Withholding modeled explicitly.** For PAW-Type-B, the adversary splits its hashpower: part mines honestly on its own, part joins a pool and submits partial proofs of work but withholds full proofs of work, then reallocates hashpower after finding one. The resulting fork race is resolved explicitly with the network-influence parameter γ rather than assumed away.
- **Numerical optimization for alternative goals.** Beyond maximizing revenue ratio, they numerically optimize alternative objective functions an adversary might pursue, such as becoming the single biggest beneficiary of the attack, or maximizing revenue gain before the difficulty adjustment arrives.

## Why This Matters

**Impact on research.** The paper supplies closed-form short-term revenue expressions for the whole L-selfish mining family, removing the need to re-run MDP models or Monte Carlo simulations for each parameter setting. It also unifies previously separate lines of work (selfish mining, intermittent mining, coin hopping, block withholding) under one efficiency metric, and it reframes the "profit lag" argument that has been used to explain why selfish mining is rarely observed in the wild.

**Real-world applications:**

- **Mining pool risk assessment:** Pool operators and members can use the PAW-Type-B results to reason about withholding exposure, particularly for smaller pools, which the paper flags as immediately threatened.
- **Miner strategy evaluation:** Miners with hardware that can switch between BTC and BCH can see that in the short term the coin-hopping gain per unit of hashpower is shared with loyal honest miners, and that the long-term comparison with selfish mining depends on the parameter regime.
- **DAA and protocol design:** Because the attacks exploit difficulty adjustment, the analysis gives protocol designers a way to reason about how responsive difficulty rules change attacker and honest-miner efficiency.
- **Security monitoring and disclosure:** The finding that non-pool honest miners can profit more than the attacker complicates detection heuristics based on "who gains," which matters for those watching for anomalous mining behavior.

**Industry relevance.** The results bear on pool design, hashpower rental and switching economics between BTC and BCH (which share a compatible Proof-of-Work algorithm, making switching seamless), and on how the security budget of a chain is interpreted when miners are free to leave.

## Future Directions

- **Interest-rate manipulation across DeFi platforms:** The authors note their framework could be applied to gains from loans and deposits on platforms whose rules depend on real time versus block count, which they leave for future work.
- **Game-theoretic extensions:** They explicitly leave for future work embedding these deviant strategies in game-theoretic frameworks with multiple strategic miners and multiple coins.
- **Re-examining the orphan-reporting claim:** The paper raises doubt about the argument that orphan reporting makes honest mining most profitable in the long run, leaving a fuller treatment of that claim open.
- **Extending PAW-Type-B:** Because the model removes the prior simplifying fork assumption, comparing its predictions against the earlier FAW/PAW models and against pool data would clarify how much the simplification mattered.

## Target Audience

Researchers and graduate students in blockchain security, applied probability, and networked-systems economics; protocol designers working on difficulty adjustment and fork choice; mining pool operators and large-scale miners evaluating strategy; and quantitatively trained analysts at funds or security firms who need to judge claims about Bitcoin incentive attacks without relying on simulation folklore.

## Authors’ abstract

Bitcoin's (BTC) Difficulty Adjustment Algorithm (DAA) has been a source of vulnerability for incentive attacks such as selfish mining, block withholding and coin hopping strategies. In this paper, first, we rigorously study the short-term revenue change per hashpower of the adversarial and honest miners for these incentive attacks. To study the long-term effects, we introduce a new efficiency metric defined as the revenue/cost per hashpower per time for the attacker and the…

[The full abstract on arXiv](https://arxiv.org/abs/2511.11538)

[Read the original paper](https://arxiv.org/abs/2511.11538)

---

Canonical: https://crypto.world/paper/incentive-attacks-in-btc-short-term-revenue-changes-and-long-term-2511.11538
