# Ark: Offchain Transaction Batching in Bitcoin

Crypto.world

> Bitcoin finalises every payment onchain, which caps throughput at roughly 10 transactions per second — about three orders of magnitude below conventional credit card systems at around 10k…

**arXiv**: 2605.20952

**Published**: 2026-05-20

**Authors**: Pim Keer, Ioannis Alexopoulos, Matteo Maffei, Marco Argentieri, Andrew Camilleri, Zeta Avarikioti

## AI summary

## Overview

**Research area:** Bitcoin Layer-2 scaling and commit-chain protocol design (cs.DC / applied cryptography), presented at the 2026 ACM SIGSAC Conference on Computer and Communications Security.

**Technical level:** Intermediate to Advanced. The paper is a full protocol specification with formal security definitions and proofs; the concepts are explained carefully, but readers will need familiarity with Bitcoin's UTXO model, Taproot scripts, and multisignature signing.

**Scope:** The paper introduces Ark, a Bitcoin-compatible commit-chain in which an untrusted operator batches offchain "virtual UTXOs" into constant-size onchain commitments, and it proves the protocol's security while measuring its onchain and offchain cost.

## What This Paper Is About

Bitcoin finalises every payment onchain, which caps throughput at roughly 10 transactions per second — about three orders of magnitude below conventional credit card systems at around 10k transactions per second. Existing offchain fixes, above all the Lightning Network and other payment channel networks, demand that users lock collateral onchain and run auxiliary protocols for rebalancing, top-ups, and routing, which keeps adoption low. Ark is proposed as the first Bitcoin-compatible commit-chain: an untrusted operator aggregates offchain transfers of virtual UTXOs (VTXOs) into succinct onchain commitments, so that users can receive funds without any onchain action and only the participants involved in a state update need to take part.

## Key Contributions

1. **The Ark protocol itself** — the first Bitcoin-compatible commit-chain with unilateral exit, opt-in fast finality, and an open architecture. Receivers can join and hold funds without issuing any onchain transaction, and funds stay usable both inside and outside Ark (external fungibility).
2. **Formal security treatment** — the authors formalise Ark and prove VTXO security, user balance security, atomicity, operator balance security, and fast finality balance security. In the process they identified two attacks against the original testnet implementation — the *hostage attack* and the *spam attack* — and fixed them by introducing *reset transactions*, which were integrated into the mainnet implementation.
3. **An experimental evaluation** — a commitment to arbitrarily many VTXOs costs a constant onchain footprint of 197 vB; unilateral exit costs ⌈log n⌉ · 150 + 107 vB for a batch of n VTXOs; and producing a commitment takes time linear in the number of users, e.g. 2.7 seconds for 200 users.
4. **A structured comparison against prior Layer-2 designs** (Table 1) covering payment channel networks, statechains, payment pools, sidechains, commit-chains such as Clique, and rollups, along axes including trust assumptions, unilateral exit, external fungibility, fast finality, parties required per state update, operator liquidity, Bitcoin compatibility, and online requirements.

## Main Findings

- **Constant onchain commitment size:** A commitment covering arbitrarily many VTXOs occupies a constant footprint of approximately 197 vB (the abstract rounds this to about 200 vB). The number of VTXOs does not change the onchain cost of the commitment.
- **Logarithmic unilateral exit:** A user exiting a VTXO without the operator's help publishes a path of virtual transactions through the batch's virtual transaction tree, costing ⌈log n⌉ · 150 + 107 vB and roughly 150 vB per VTXO for a batch of n VTXOs. The authors note this fee cost may make small VTXOs uneconomical to exit.
- **Constant cooperative exit:** With a responsive operator, exiting costs 𝒪(1) onchain transactions, adding one output per user in the next commitment.
- **Commitment time is linear in participants:** Producing a commitment scales linearly with the number of users involved; 200 users require a mere 2.7 seconds.
- **Two concrete attacks on the testnet implementation:** The authors found a **hostage attack** and a **spam attack**, responsibly disclosed them, proposed fixes based on reset transactions, and the fixes are now in the mainnet implementation.
- **Security reduces to two properties:** VTXO security (an unexpired VTXO committed in an onchain batch can only be spent with a valid witness, and an honest holder can always redeem it onchain before batch expiry) plus Ark atomicity (every action — boarding, transaction, batch swap, exit — either consumes all inputs and creates all outputs, or does neither). Balance security for users and for the operator follows from these.
- **Delayed finality by default, fast finality by opt-in:** Under a malicious operator, Ark gives delayed finality — a payment is final only once its commitment is confirmed onchain and survives the dispute period. Under a *rational* operator plus an *existentially honest* signer committee (at least one honest member), users can opt in to spend VTXOs immediately without risk of reversion.
- **Online requirements are asymmetric:** Only the operator must stay continuously online; users need only connect periodically to refresh VTXOs, entirely offchain. Users making fast finality payments must stay online until their funds are spent.
- **Interactivity is limited to involved parties:** Unlike Clique-style designs, which need signatures from all users to update state, Ark needs only the operator plus the users involved in that specific update.
- **Covenants are emulated, not required:** Because Bitcoin lacks script-level covenants, the virtual transaction tree's spend restriction is enforced with n-of-n multisignatures (e.g. MuSig2), or more efficiently by having each holder sign only along the path to their own VTXO. This reduces interaction without harming safety, since every honest holder must approve the transactions needed to redeem their VTXO.

## Methodology in Plain English

The authors design a protocol from Bitcoin's existing primitives and then prove and measure it.

The building blocks are three. **VTXOs** are offchain outputs locked by a Taproot script with an unspendable key path and two script paths: a collaborative path requiring the operator's signature, and a unilateral path that needs only the user's signature after a relative timelock. **Batches** are onchain outputs, funded mostly by the operator's own liquidity, that can be swept back by the operator after an expiry height or "unrolled" — split into individual VTXOs — via a presigned binary tree of virtual transactions that optimistically never touches the chain. **Connectors** are outputs that spawn *anchor* outputs, which atomically bind an offchain state change to a future onchain commitment.

Payments work offchain: Alice signs a virtual transaction spending her VTXO and asks the operator to co-sign, producing a VTXO for Bob. Because a virtual transaction can in principle be double-spent, Bob finalises by *batch swapping*: he forfeits the received VTXO in exchange for a fresh one in the next commitment, signing a forfeit transaction that spends both the old VTXO and an anchor with SIGHASH_ALL. The forfeit is valid only if the commitment containing the anchor is onchain, which gives atomicity — either Bob gets his new VTXO or the operator gains nothing.

Joining (*boarding*) is a two-step process: the user posts an onchain output spendable jointly with the operator or unilaterally after a timelock, and the operator then consumes it in the next commitment. Exiting is either collaborative (a batch swap that pays out a plain onchain UTXO instead of a new VTXO) or unilateral (the user broadcasts the path through the virtual transaction tree themselves).

For the security analysis, the authors work in a UTXO ledger model with a robust ledger parameterised by depth k and wait u, a synchronous network with known delay Δ, EUF-CMA signatures and collision-resistant hashing, and a Byzantine adversary that can corrupt any subset of parties including the operator, control scheduling and the mempool, but cannot break cryptography. Fast finality adds a rational-operator and existentially honest committee assumption. The evaluation measures onchain footprint in virtual bytes and commitment production time.

## Why This Matters

Ark separates two things that Layer-2 designs usually conflate: the ability to *receive* money and the requirement to *lock* money. That lowers the barrier to entry for new users and, because VTXOs remain spendable to anyone including non-users and can be moved back onchain, avoids fragmenting liquidity across channels.

**Impact on research:** The paper supplies a formal, proven commit-chain construction that needs no new opcodes and no covenant support on Bitcoin, and it feeds back into practice — the two attacks it found were fixed in a live mainnet implementation. Its constant-commitment / logarithmic-exit cost profile gives a new data point for the trade-off space in Layer-2 comparison work.

**Real-world applications:**

- Instant retail and point-of-sale Bitcoin payments, where a merchant can accept funds without opening a channel or pre-funding anything.
- Exchange or custodial deposit/withdrawal batching, amortising many user payouts through one constant-size onchain commitment.
- Payroll and airdrop-style disbursements to many recipients at once, with one output per user only on cooperative exit.
- A settlement substrate on which other Layer-2 protocols can be set up, since Ark funds are externally fungible and can be moved out to onchain Bitcoin.

**Industry relevance:** The operator funds commitments from its own liquidity and can charge fees via fee VTXOs, giving a clear business model for infrastructure providers, while users retain a unilateral exit guarantee against an adversarial operator. The integration of the disclosed fixes into mainnet signals near-term deployability rather than a purely theoretical result.

## Future Directions

- **Covenant support on Bitcoin:** The authors note that native covenants would reduce interactivity compared with n-of-n multisignature emulation. Quantifying how much signing round-trips and latency this would save — and whether covenant-based Ark variants remain as safe — is a natural follow-up.
- **Robustness of the fast finality assumption:** Fast finality rests on a rational operator and an existentially honest signer committee. How such a committee is selected, incentivised, and kept live in practice, and what happens if rationality assumptions fail, remains open.
- **Economic viability of small VTXOs:** The paper observes that the onchain fees for a logarithmic-depth unilateral exit may make small VTXOs uneconomical to exit. Fee management, VTXO consolidation policies, and dust handling are open engineering questions.
- **Fuller evaluation and deployment data:** The truncated content reports commitment footprint, exit cost, and commitment production time (2.7 seconds for 200 users); broader measurements of throughput, operator liquidity requirements, and behaviour under adversarial operators are not reported here and would be the obvious next empirical step.

## Target Audience

Researchers and protocol engineers working on Bitcoin scaling and Layer-2 design; cryptocurrency engineers evaluating commit-chains, payment pools, or channel factories as deployment options; and security analysts interested in formal proofs of offchain protocols and in the concrete attack classes (hostage and spam) that arise in operator-assisted batching. Readers without a distributed-systems or Bitcoin scripting background will find the formal sections demanding, but the protocol overview and evaluation are accessible to practitioners who know the UTXO model.

## Authors’ abstract

Bitcoin is the cryptocurrency with the largest market capitalisation, but its widespread adoption is fundamentally limited by the scalability constraints of its consensus algorithm, which requires every transaction to be confirmed onchain. To address this, several Layer-2 scalability solutions have been proposed to move payments offchain -- most notably, the Lightning Network. However, their deployment remains hindered by cumbersome setup requirements: users must lock funds…

[The full abstract on arXiv](https://arxiv.org/abs/2605.20952)

[Read the original paper](https://arxiv.org/abs/2605.20952)

---

Canonical: https://crypto.world/paper/ark-offchain-transaction-batching-in-bitcoin-2605.20952
