# APoW: Auditable Proof-of-Work Against Block Withholding Attacks

Crypto.world

> Under standard proof-of-work, a miner who finds a valid block can simply hide it while continuing to submit valid low-difficulty shares, so the pool sees statistically normal behaviour but earns less.

**arXiv**: 2601.02496

**Published**: 2026-01-05

**Authors**: Sergio Demian Lerner

## AI summary

## Overview

**Research area:** Proof-of-work consensus design and mining-pool security, specifically countermeasures against block withholding attacks (BWAs) in Bitcoin-style cryptocurrencies.

**Technical level:** Advanced. The paper assumes familiarity with nonce searching, share difficulty, mining-pool payout schemes (PPS, FPPS, PPLNS, PPS+), block templates, and ASIC mining hardware.

**Scope:** A construction-paper introducing Auditable Proof-of-Work (APoW), a Hashcash-style PoW variant that lets miners probabilistically prove after the fact which nonce regions they searched, together with a security, incentive, and Bitcoin-deployment analysis. The paper reports no experimental results or benchmarks.

## What This Paper Is About

Under standard proof-of-work, a miner who finds a valid block can simply hide it while continuing to submit valid low-difficulty shares, so the pool sees statistically normal behaviour but earns less. Because block discovery is a rare, low-probability event, an absence of blocks is not evidence of misbehaviour, and pools cannot distinguish malice from variance or produce cryptographic proof of withholding. The paper's goal is a PoW construction in which unsuccessful search effort leaves an auditable trace, so that other miners can later re-scan a miner's claimed nonce ranges and probabilistically detect withheld blocks, without trusted hardware, trusted third parties, or (optionally) without any consensus change at all.

## Key Contributions

1. **The APoW construction itself.** A dual-predicate PoW in which a "v-miner" simultaneously searches for hashes matching an unpredictable *d*-bit pattern `b2` (yielding ordinary shares or full blocks, credited normally) and hashes matching a *d*-zero-bit pattern `b1` (used only as auditing evidence). This lets auditing happen concurrently with productive mining rather than as separate, unpaid re-mining.
2. **A formal notion of auditable work units.** A miner's search space `S_i^A` for round `i` is a set of tuples `(G, n_s, n_e)` — a block template header plus a contiguous nonce interval — all sharing the same parent block. Pool-side rules are specified for constructing these units, including disabling timestamp (nTime) rolling and auditing only up to the last share submitted for a given template and timestamp.
3. **Two block designs and a fork-limiting rule.** Scheme 1 uses non-empty v-blocks that are sealed with a signature over `(i, G_i, n, j, G_j)` and must be included immediately after the audited block; Scheme 2 uses empty v-blocks that need no sealing because they are already unique. Any v-block must be immediately followed by a regular PoW block, consecutive v-blocks are disallowed, and auditing is permitted up to a maximum depth `D` rounds after the audited block.
4. **Three v-block attribution mechanisms.** (a) Pay to `poolAddress_i` embedded in the template, optionally with a `lastBlock_i` field so rewards are split as in the referenced block; (b) an `auditorsRoot` Merkle root of eligible auditor public keys, with the block ID at height `i` seeding a CSPRNG to select the auditor and a Merkle inclusion proof in the v-block; (c) a zero-knowledge proof of knowledge of the message `M` revealing everything except the nonce, so only the finder can claim the v-block.

## Main Findings

- **Payment-scheme exposure classification.** Block-conditioned schemes (PPLNS, score-based) put variance on the miner and are rated Strong resistance with no BWA profitability. PPS+ is Shared variance, Moderate resistance, Partial profitability. PPS is Pool variance, Weak resistance, profitable. FPPS is Pool variance, Very weak resistance, profitable.
- **Estimated share of hashrate at risk.** Based on public pool disclosures, historical hashrate distribution, and advertised payout options, the author estimates approximately **25%–35% of the current Bitcoin network hashrate** participates in pools offering PPS or FPPS-style payouts and is therefore economically susceptible to BWAs absent auditing. The estimate excludes pools relying solely on PPLNS or score-based methods and is stated to vary over time.
- **Hash-caching is economically prohibitive.** A miner performing **200 TH/s** run for **600 seconds** until a block is found would need to store about **4 exabytes (2^62 bytes)** of 256-bit digests; such a machine would cost more than **10 Trillion dollars as of 2025**. Storing only **16 bits per digest** allows **8 hash results per RAM access** (assuming a **128-bit wide data bus**), an **8X** speedup, but the machine still costs over **1T dollars**. Storing **1 bit per hash** in a **128-bit hint word** with a custom overclocked ASIC still costs **39B USD** because of RAM requirements.
- **Limited ASIC compatibility.** The scheme is stated to be incompatible with the majority of Bitcoin ASICs, because Bitcoin difficulty 1 requires checking **32 leading zeros**, and requiring 32 zero leading bits for v-mined hashes would make caching economically viable. If ASICs checked only **20 leading zero bits** and partial matches were transferred quickly to a host CPU, caching RAM would still cost about **10M USD per machine** — but the author states this possibility has not been tested on actual ASICs. The cleanest option is a small circuit change letting the ASIC compare leading bits against a dynamically loaded pattern instead of only zeros.
- **BWA during v-mining is largely neutralised by rotation**, since auditor roles would rotate and only a small portion of hashrate goes to auditing. For stronger guarantees, the paper proposes **recursive auditing**: v-blocks auditing other v-blocks, accepted and rewarded up to a maximum audit depth **D2** (`0 < j-k ≤ D2`), with the ASIC comparison circuit modified to check two leading bit patterns, where `b1` matches the pattern `b2^A` that miner A used at height `k`.
- **Auditor collusion is bounded by delayed payouts.** A v-miner colluding with a withheld miner is not a high risk because detection is probabilistic and the pool has many auditing opportunities; the pool may retain rewards for a predefined number of blocks (the paper's example is **100 blocks**) to act as a stake that can be forfeited.
- **Prior approaches have structural drawbacks.** Rosenfeld's oblivious shares require a hard fork with full-node and light-client upgrades, introduce share-submission latency versus immediate Stratum v2 block broadcast, and concentrate authority in a pool holding secret information. Rosenfeld's "pop quiz" and peer-review audits are often distinguishable from real work (e.g. via stale block height or unexpected parent hash), consume hashpower that earns nothing, and are too weak to gather evidence before a pool goes bankrupt.
- **APoW is compatible with oblivious shares**, allowing centralized pools to use oblivious shares and decentralized pools to use APoW while competing fairly.
- **Useful without consensus change.** Because auditing shares are indistinguishable from ordinary shares, a pool can fund audits from existing reserves under normal payout rules, as long as the expected auditing cost is below the expected loss from a successful BWA.

## Methodology in Plain English

This is a design-and-analysis paper rather than an empirical one. The author starts from the observation that PoW lets anyone verify *successful* work but leaves *unsuccessful* work invisible, then constructs a PoW puzzle with two winning conditions at once: one that behaves like normal mining (paid shares and blocks) and one that serves only as a receipt that a specific nonce range was scanned. The receipt-producing pattern `b2` is derived by hashing the block ID at height `j-1`, so a miner cannot precompute it or grind templates in advance — a fairness measure. The author then specifies exactly how a re-scanning miner packages its proof into a "v-block", how the chain limits the damage from free v-block equivocation (immediate-inclusion rule, no consecutive v-blocks, bounded audit window), and how the reward reaches the right party (pool address, auditor Merkle root with CSPRNG selection, or zero-knowledge proof). Finally, the paper works through the obvious ways an adversary might cheat — caching hashes in RAM, withholding during auditing, colluding with the audited miner — and estimates costs using arithmetic on hash rate, digest size, memory bandwidth and electricity, rather than running experiments. Deployment feasibility is assessed qualitatively against Bitcoin's difficulty-1 target, ASIC pipeline design, Stratum pool conventions, and hardware lifetimes of **2-3 years**.

## Why This Matters

**Impact on research.** The paper reframes BWA defence from a statistics problem ("too few blocks, probably bad luck") into an accountability problem with cryptographic evidence, and it does so with an orthogonal auditability layer that preserves public verifiability and difficulty adjustment. Its classification of payout schemes by BWA exposure, and the notion of auditable work units, give other researchers concrete objects to formalise and attack.

**Real-world applications:**

- Decentralized mining pools that pay workers directly from the coinbase transaction via multiple outputs, using `lastBlock` or `auditorsRoot` attribution, without requiring KYC or reputation.
- Pool-level auditing deployed today with no consensus change, including "verifiable pay-for-auditing" funded from existing pool reserves.
- Risk management for PPS/FPPS pools, which bear all variance and are described as the most BWA-exposed category.
- Smart-contract or sidechain-mediated reward distribution, where `poolAddress` is controlled by a contract or a sidechain exit address.

**Industry relevance.** The estimate that roughly **25%–35% of Bitcoin hashrate** sits in PPS/FPPS pools quantifies a live economic exposure for pool operators. The hardware analysis also makes the trade-off concrete for ASIC vendors: a small change to the compare circuit (match a loaded bit pattern, not just zeros) is the cleanest path, while full incompatibility with existing ASICs and the need for a hard fork are stated plainly as adoption costs.

## Future Directions

- **Empirical ASIC validation.** The paper explicitly states that the **20-leading-zero-bit** partial-match variant, which would make APoW workable on some existing hardware, "has not been tested" on actual ASICs. Testing that, and prototyping the dynamic bit-pattern compare circuit, are the obvious next steps.
- **Mining fairness and selfish-mining analysis of v-blocks.** The author notes that because a v-block's parent is chosen post-PoW, a v-block could be sealed to fit different chain slots, and that "a careful analysis is required including how this impacts mining fairness and selfish mining attacks." No such analysis is provided here.
- **Formalising the zero-knowledge attribution variant.** The paper defers a "generalized formalization of a ZK-PoW" to Appendix A, leaving the concrete proof system, its overheads and its soundness as open work.
- **Deployment and incentive design for a dual-PoW hard fork.** The paper sketches a fork in which APoW and existing SHA-256 mining co-exist and pools add extra incentives for v-mining, arguing the **2-3 year** hardware lifetime is enough for most infrastructure to adapt. How such a fork is coordinated, and whether recursive auditing at depth **D2** is enough to keep residual BWA damage negligible, remain open questions.

## Target Audience

Mining-pool operators and payout-scheme designers evaluating BWA exposure; Bitcoin and consensus protocol researchers interested in PoW variants and pool decentralisation; ASIC architects assessing the cost of adding pattern-matching to the hash comparator; and cryptoeconomics researchers studying incentive compatibility in pooled mining. Readers need prior familiarity with Stratum-style pool protocols and proof-of-work mechanics to follow the construction; the security-cost arguments are accessible to a broader technical audience.

## Authors’ abstract

We introduce Auditable Proof-of-Work (APoW), a novel proof-of-work (PoW) construction inspired by Hashcash-style nonce searching, which enables the auditing of other miners' work through accountable re-scanning of the nonce space. The proposed scheme allows a miner to probabilistically attest to having searched specified regions of the nonce space in earlier mining rounds, while concurrently earning rewards for performing productive work for a new block or pool share. This…

[The full abstract on arXiv](https://arxiv.org/abs/2601.02496)

[Read the original paper](https://arxiv.org/abs/2601.02496)

---

Canonical: https://crypto.world/paper/apow-auditable-proof-of-work-against-block-withholding-attacks-2601.02496
