Research
Proof of Source of Funds: Efficient On-chain Provenance of Cryptoassets
Today, crypto compliance works retroactively: centralized blockchain-intelligence firms continuously analyze the public transaction graph to taint or block certain funds, and platforms absorb that…

- arXiv
- 2606.10172
- Published
- 2026-06-08
- Authors
- Alireza Kavousi, István András Seres, Zhipeng Wang
AI summary
Overview
Research area: Applied cryptography and blockchain systems — specifically zero-knowledge proof systems, on-chain compliance (AML/KYC), and cryptoasset provenance/tracing.
Technical level: Advanced. The paper assumes familiarity with zero-knowledge succinct arguments of knowledge, Incrementally Verifiable Computation (IVC), folding schemes, Merkle commitments, and the account-based vs. UTXO ledger models.
Scope (one sentence): The paper designs, formalizes, and benchmarks Proof of Source of Funds (PoSoF), a user-side zero-knowledge protocol that proves a crypto deposit originates only from a pre-approved set of compliant sources, verified on-chain in constant time without revealing the transaction path.
What This Paper Is About
Today, crypto compliance works retroactively: centralized blockchain-intelligence firms continuously analyze the public transaction graph to taint or block certain funds, and platforms absorb that cost and risk. The authors argue this is fragile (attackers route funds through fresh addresses), inaccurate (false positives and negatives), invasive (honest users lose financial privacy), and vulnerable to "tainting attacks" where anyone can send illicit dust to an honest address. Their goal is to invert the model: instead of the platform tracing the user, the user locally builds a zero-knowledge proof that their funds descend exclusively from compliant origins, and the platform only runs a constant-time check.
Key Contributions
-
Formalization of prover-side provenance. The paper redefines compliance as user-side provenance rather than platform-side surveillance, and introduces a unified temporal DAG abstraction in which nodes are
(address, sequence-coordinate)pairs. This single model covers both account-based ledgers (e.g., Ethereum) and UTXO ledgers (e.g., Bitcoin) as a generalized value-flow model, backed by formal security definitions of the PoSoF statement and witness. -
Efficient IVC-based protocol instantiation. The authors construct a PoSoF protocol using Incrementally Verifiable Computation (via modern folding schemes such as Nova) so that a per-edge step circuit and a nodal merge circuit are folded incrementally while the prover walks the DAG, yielding a constant-size proof and
O(1)on-chain verification. The proof hides intermediate transaction topology, intermediary addresses, and which specific compliant origins were used. -
A rigorous predicate set over the provenance sub-DAG. The circuit enforces six predicates intended to be jointly exhaustive against known attack vectors: on-chain integrity and distinctness (no reusing a transaction across branches, plus fractional capacity
v_e ≤ v(tx)), source anchoring (every source node proves membership in the compliant set), edge linkage (sender/receiver address continuity to stop topology forgery), temporal ordering (stricth(e_in) < h(e_out)to stop "time-travel" exploits), flow solvency (V_out(a) ≤ V_in(a)), and terminal execution binding (proof is non-malleable and tied to the exacttx_target). -
Implementation plus application-layer anti-double-spend defenses. The paper reports an Ethereum-compatible proof-of-concept and proposes public nullifiers and private change commitments so a user cannot "double-spend" the same clean history across multiple deposits.
Main Findings
- Proving is practical on consumer hardware. The paper's benchmarks report roughly 1.8 s to incrementally update a user's PoSoF proof per new transaction via IVC.
- On-chain verification is constant-time. Final EVM verification costs about 1.5 ms, or roughly 800k gas, regardless of how large the compliant sub-DAG is.
- Fractional provenance is supported without inflation. The capacity variable
v_e ≤ v(tx)lets a user claim only part of a mixed transaction as compliant, while flow solvency (∑v_out ≤ ∑v_in) prevents claiming more compliant volume than the ledger records. - Tainting attacks become irrelevant by construction. Because only a compliant sub-DAG is proven, an honest user who receives illicit funds simply excludes that edge from their witness; the paper illustrates this with a complex-DAG scenario where a non-compliant inflow of 500 is omitted from the proven graph.
- Binary allow/blocklists are argued to fail. The authors contend that inclusion/exclusion-list designs are evadable via fresh addresses, susceptible to dust-based tainting, and require on-chain verifiers to synchronize with constantly changing lists — costs a static compliant-origin anchor avoids.
- Compliance cost context is quantified. The paper cites industry figures of over $206 billion spent annually by financial institutions on financial-crime compliance, AML operations consuming up to 19% of a firm's annual revenue, and baseline KYC/AML overhead averaging roughly $620,000 per year for small-to-mid-sized crypto platforms, scaling into the tens of millions for major exchanges.
Methodology in Plain English
The user starts from their on-chain history (the global ledger graph) and searches backwards for a sub-graph of transactions that leads only to sources on an approved list — for example, regulated exchanges where they already passed KYC. They may take only a fraction of each transaction's value, because a single transfer can mix clean and dirty money.
They then prove, in zero knowledge, that this sub-graph obeys a fixed set of rules: every edge is a real, finalized transaction on the canonical chain; every starting point really is on the approved list; each hop's recipient is the next hop's sender; transactions strictly move forward in time; no address passes on more value than it received; no transaction is reused twice; and the final amount covers the specific deposit being authorized.
Generating a single giant proof for the whole graph would be too expensive, so the paper uses Incrementally Verifiable Computation. The user proves one edge at a time and folds each partial proof into a running accumulator, merging branches where the graph joins. The resulting constant-size proof is posted on-chain and checked by a smart contract in a fixed amount of time.
On the application side, the paper sketches additions such as public nullifiers (so a clean-history proof cannot be spent twice) and private change commitments, and it illustrates the approach on three worked examples: a single path, a multi-source merge where two inflows of 80 and 50 fund a 100 transfer, and a more complex graph mixing compliant and non-compliant branches.
The evaluation is described as an Ethereum-compatible prototype; the truncated content does not report the hardware specification, the number of benchmark trials, proof sizes, or dataset dimensions for the experiments.
Why This Matters
Impact on research. The paper reframes compliance from a surveillance problem into a proof problem, and it supplies a formal model (temporal DAG plus six predicates) that other privacy-protocol designers can build on. If the security arguments hold, it separates two goals that prior work treated as coupled — hiding the transaction graph and demonstrating that funds are clean — and it gives a concrete, benchmarked cost for doing so.
Real-world applications (as described in the paper):
- Compliant privacy pools. Deploy PoSoF as a strict admission-control check on deposits to pools such as Tornado Cash, so the pool's anonymity set stays intact while its liquidity is guaranteed free of illicit funds — responding directly to the 2022 US Treasury sanctioning of Tornado Cash.
- Proactive AML and exchange onboarding. Exchanges require a provenance proof alongside the deposit transaction and perform
O(1)verification before crediting funds, instead of paying for retroactive chain surveillance and discovering problems after money has entered. - Privacy-preserving undercollateralized lending. A borrower proves a sustained, cryptographically verified history of legitimate cash flows, letting a protocol treat proven historical liquidity as a credit signal and reduce structural overcollateralization (the paper cites 150% as typical) without revealing net worth or counterparties.
- Regulated DeFi gatekeeping more broadly. Any protocol that must exclude sanctioned or illicit capital can use provenance as an admission rule while keeping its own users' histories opaque to the public.
Industry relevance. The paper targets the economics of blockchain analytics and AML operations directly: it moves the compute and analysis burden from platforms and intelligence vendors to the individual prover, and it replaces a continuously synchronized blocklist with a comparatively static anchored allowlist. The reported ~800k gas verification cost is the practical number a protocol team would weigh against existing compliance spend.
Future Directions
- Extending beyond the restricted setting. The formal model restricts intermediate temporal nodes to externally owned accounts (EOAs) and assumes a single asset type such as native ETH; generalizing to contract accounts and multi-asset flows is left open.
- Formal security analysis of the full system. The paper states security properties formally, but the truncated content does not report a full reduction-based proof or an independent audit of the deployed prototype and its nullifier/change-commitment defenses.
- Governance of the compliant set. The design anchors trust in the set
Cof approved origins; how that set is curated, who may update it, and how users challenge exclusion are questions the paper's framing raises rather than resolves. - Prover cost at scale. The reported ~1.8 s per-transaction update is for the prototype; how this grows with long histories, many branches, or weaker consumer devices (and whether proofs can be delegated without leaking history) is not reported.
Target Audience
Researchers and protocol engineers working on zero-knowledge systems, privacy-preserving cryptocurrencies, and blockchain compliance tooling. It will also interest compliance and risk teams at exchanges and DeFi protocols who need to understand what user-side provenance proofs can and cannot replace, and regulators or policy researchers evaluating whether proof-based compliance is a credible alternative to surveillance-based tracing. Readers without a background in succinct arguments and IVC will find the protocol sections demanding.
Authors’ abstract
Regulatory compliance is increasingly mandatory for decentralized finance and privacy-enhancing technologies. Current approaches rely on binary inclusion/exclusion lists or retroactive graph analysis by centralized blockchain intelligence firms. This approach strips honest users of their financial privacy, leads to false positives and negatives, and forces decentralized platforms to bear the burden of on-chain transaction monitoring. In this work, we propose a paradigm…
The full abstract on arXiv