Mixers, tumblers, and CoinJoin

As of August 2026: Roman Storm has not been sentenced on his August 2025 conviction; his motion for acquittal remains undecided, and retrial on the hung money-laundering and sanctions counts is scheduled for April 2027.

Also called: tumblers · CoinJoin.

What is a crypto mixer?

Every bitcoin ever moved is still visible. The blockchain is a public ledger, and anyone with the right tools can follow a coin from the wallet that received a ransom payment, through every subsequent transfer, to the exchange account where someone finally tried to cash out. That transparency is the foundation of modern crypto investigation, and mixers exist to attack it.

A mixer (or tumbler) pools coins from many users and pays out equivalent value from the pool. Your coins go in; different coins, or coins whose history now points to a crowd rather than a person, come out. The layering logic is identical to running money through a chain of shell accounts, except the whole thing happens in software and takes minutes.

Three designs dominate. Custodial tumblers (Bitcoin Fog, Helix, ChipMixer) were businesses: an operator took deposits, charged a fee, and paid out from a common pool. Smart-contract mixers like Tornado Cash removed the operator: an immutable contract on Ethereum accepts fixed-size deposits and lets each depositor later withdraw to a fresh address, with cryptography hiding which deposit funded which withdrawal. CoinJoin wallets (the model behind Samourai’s Whirlpool) are collaborative: many users co-sign one large transaction with equal-sized outputs, so nobody ever takes custody and no observer can say whose input became whose output.

The designs matter legally (courts have treated custodial operators very differently from immutable code), but to an investigator tracing stolen funds, they present the same problem: the transaction graph stops telling the truth.

How does mixing work?

1. A user holding traceable coins (hack proceeds, darknet revenue, or simply funds they want private) sends them to the mixer, usually in standard denominations so every deposit looks alike.
2. The coins join a pool alongside deposits from dozens or thousands of other users. Pool size is the product’s real feature: the bigger the crowd, the weaker any inference about who owns what.
3. After a delay, the user withdraws equivalent value (minus fees) to brand-new addresses with no history. Custodial services paid out other people’s coins; Tornado Cash used zero-knowledge proofs so even the contract couldn’t link deposit to withdrawal.
4. The withdrawn funds then move on, often into chain-hopping swaps or exchange deposits, with their visible history now beginning at the mixer’s edge rather than at the crime.

Why does mixing work, and where does it fail?

Blockchain analysis rests on one assumption: flows can be followed. Clustering heuristics group addresses controlled by the same actor; tracing tools walk value from a theft to a cash-out point. A mixer attacks the assumption directly. After a well-populated mix, the honest answer to “whose coins are these?” is a probability spread across every participant in the pool. The trail doesn’t vanish; it becomes ambiguous, and ambiguity is often enough, because an exchange compliance officer cannot freeze a probability.

Mixers also solve a timing problem for thieves. A publicized hack starts a race: exchanges and stablecoin issuers blacklist the attacker’s addresses within hours. Coins parked in a mixer wait out the alert, then emerge from addresses no blacklist has ever seen.

But mixing has structural weaknesses. Pools are finite: moving a nine-figure theft through a mixer means dominating the pool, and when most of the coins in a mix are yours, mixing with yourself hides little. Researchers have repeatedly demixed transactions using amounts, timing, and address-reuse mistakes. And the mixer itself becomes the chokepoint: every major tumbler’s operators kept records, made mistakes, or touched infrastructure that investigators could seize.

The Tornado Cash saga: code, sanctions, and a half-decided verdict

No mixer has tested the law like Tornado Cash. Launched in 2019 as a set of immutable smart contracts on Ethereum, it became the venue of choice for the biggest thieves in crypto, including North Korea’s Lazarus Group, which pushed hundreds of millions in hack proceeds through it. In August 2022, OFAC added Tornado Cash to the SDN sanctions list: the first time the US had sanctioned a decentralized protocol rather than a person or company. Overnight, interacting with the contracts became legally radioactive for US persons.

Then the courts pushed back. In Van Loon v. Department of the Treasury, the Fifth Circuit ruled in November 2024 that Tornado Cash’s immutable smart contracts are not the “property” of any foreign national under the sanctions statute, so OFAC had exceeded its authority. In March 2025, Treasury delisted Tornado Cash, framing the move as its own discretion rather than a concession, which triggered a further fight over whether it was evading a final judgment.

The criminal case against developer Roman Storm split the difference. In August 2025, an SDNY jury convicted Storm on one count: conspiracy to operate an unlicensed money transmitting business. On the heavier counts (money-laundering conspiracy and sanctions-evasion conspiracy) the jury hung. Prosecutors moved to retry him; after scheduling shifts, the retrial is set for April 2027. Storm’s motion for acquittal, argued in April 2026, remains undecided, and he has not been sentenced. The deepest question the case raises (when does publishing privacy code become running a criminal business?) is, as of August 2026, still open.

The operators of custodial mixers have fared far worse. Roman Sterlingov, who ran Bitcoin Fog for a decade as it moved about $400 million (including funds from the Bitfinex hack laundering), got 12 years and 6 months in November 2024. Helix’s Larry Harmon, who cooperated extensively, got 3 years and forfeited over $400 million. ChipMixer, which the DOJ says processed over $3 billion in unlawful transactions, was dismantled in a March 2023 US–German–Europol operation; its alleged operator remains on the FBI’s wanted list. And Samourai Wallet’s founders, whose Whirlpool CoinJoin service the government said laundered over $237 million, pleaded guilty in July 2025 and were sentenced in November 2025 to 5 and 4 years, with nearly $238 million forfeited.

How mixers get caught

Enforcement almost never breaks the cryptography. It breaks everything around it.

The operators. Sterlingov was identified partly through decade-old registration trails from Bitcoin Fog’s earliest days. ChipMixer fell when German police seized its back-end servers, along with $46 million in crypto. Custodial mixers concentrate risk in one place: whoever runs the pool can be found, and their records unlock the pool’s history.

The edges. Coins must enter a mixer from somewhere and exit to somewhere, and both edges are visible on-chain. Analytics firms maintain live clusters of mixer addresses, so “funds arrived from a mixer” is itself a detection signal. Exchanges, the choke point where crypto becomes money, score mixer exposure as high risk, freeze deposits, and file reports that reach investigators through the systems described in detection.

The math. Demixing research exploits denominations, timing correlations, and user error. A thief who deposits an unusual amount, withdraws on a predictable schedule, or later consolidates “clean” outputs into one wallet has undone the mix themselves.

The law. Where tracing stalls, prosecutors have simply charged the service. Every major custodial mixer of the last decade (Bitcoin Fog, Helix, ChipMixer, Samourai) has ended in conviction, takedown, or an FBI wanted poster, on theories that don’t require proving where any particular coin went: operating an unlicensed money transmitting business, and forfeiture of everything the service touched. The mixer’s promise is that the pool protects the user. Its record is that nothing protects the pool.

▶ Play video

First published August 28, 2026 · Last updated August 28, 2026 · Legal statuses current as of August 2026