PillarPrivacy

Bitcoin Privacy 101: CoinJoin and Beyond

~11 min read

A CoinJoin merges several people's payments into one transaction so that no observer can tell which output belongs to which input. It attacks exactly one assumption, the one chain-analysis firms lean on hardest: that every input to a transaction has the same owner. It does not hide amounts, it does not hide timing, and it does not survive careless spending afterwards.

That last sentence is the part most guides skip, and it is the reason a lot of people who thought they had privacy did not have any. The technique is sound. The way it gets used is usually where it falls apart.

Two transactions side by side. On the left an ordinary transaction with three inputs of 0.31, 0.07 and 0.62 BTC feeding a 0.85 BTC payment and 0.1497 BTC change, annotated with an analyst's conclusion that all three inputs are one wallet. On the right a CoinJoin where three separate people contribute one input each and receive three identical 0.1 BTC outputs plus three unequal change outputs.
The three equal outputs on the right are interchangeable. The grey change outputs underneath them are not, and they are still tied to the input that produced them.Original diagram by BTCLinks.

What does a CoinJoin actually do?

Greg Maxwell posted the idea to Bitcointalk on 22 August 2013, and the whole thing rests on one property of the protocol that he put in bold at the time: the signatures inside a transaction are completely independent of each other. Nothing in Bitcoin says one person signed them all. That was always just an assumption analysts made because it is usually true.

So a group can agree on a set of inputs to spend and a set of outputs to pay, then each person signs only their own input, and the pieces get merged into one valid transaction. Maxwell's own description of the result is the important bit: such a transaction is "externally indistinguishable from a transaction created through conventional use."

The Bitcoin wiki calls the assumption being attacked the common-input-ownership heuristic, and describes it as "unreasonably effective" in practice. Break it and a large amount of clustering work stops being reliable. That is the entire product.

What a CoinJoin hides and what it leaves visible
SignalHidden by CoinJoin?Why
Which output belongs to which inputYesEqual-valued outputs are interchangeable by construction
That the inputs share an ownerYesEach participant signed only their own input
The amounts involvedNoEvery value is written on chain in the clear
The change outputsNoChange is whatever is not equal-valued, so it is obvious
That a CoinJoin happened at allNoEqual-output transactions are a recognisable pattern
Your IP addressNoSeparate problem, separate fix (Tor, your own node)
What you do with the coins nextNoThis is where most of the privacy is actually lost
Read the bottom three rows twice. A CoinJoin is one transaction in a history, and the transactions on either side of it are the ones that give you away.Table by BTCLinks, drawing on the privacy analysis maintained on the Bitcoin wiki, verified 2026-07-29.

Where does it leak?

Amounts are the first hole. The wiki's framing is blunt: value goes in, value minus fee comes out, and an adversary can simply search the chain for output amounts that match. Equal-valued outputs exist precisely to defeat that search, which is why serious implementations use them and why the fixed denominations feel so awkward to use.

Change is the second hole, and it follows from the fix for the first. If the anonymous outputs are the equal ones, then the odd-sized leftovers announce themselves. Every participant walks out of a CoinJoin holding one clean output and one output that is still fully attached to their past.

The third hole is you. This is the one that matters most and gets discussed least.

Diagram showing three mixed 0.1 BTC outputs each with an anonymity set of 100. On the upper path all three are spent in one later transaction to an exchange, which re-links them. On the lower path each is spent separately on different days to different destinations, and no link is re-created.
The mix bought each of these an anonymity set of one hundred. The upper path spends it all in a single transaction, and the common-input assumption applies again.Original diagram by BTCLinks.

Spend three mixed outputs together and you have signed them all in one transaction, which is the exact pattern CoinJoin exists to break. Combine a mixed output with an unmixed one and it is worse: you have handed over both histories and welded them together. Send the result straight to an exchange account in your legal name and the whole exercise was theatre.

Timing helps analysts too. So does using a wallet whose fee-estimation habits and script-type choices produce a recognisable fingerprint, which is a documented technique for spotting change outputs. And if you connect to somebody else's node over a plain connection while doing any of this, the network layer can leak what the chain layer did not. Running your own full node removes that entire category of problem and costs about the price of a used laptop.

Which CoinJoin tools are actually running in 2026?

This is where most articles on this subject are quietly out of date, because the answer changed twice in eighteen months and not in the direction anybody expected.

Wasabi Wallet. The software is alive and maintained. Version 2.8.1 was published on 22 July 2026 and added Taproot addresses by default. What ended is the coordinator. zkSNACKs, the company that built Wasabi, discontinued its coinjoin coordination service on 1 June 2024. As of 29 July 2026 the Wasabi site ships no default coordinator at all and tells users to "find a third-party Wasabi Coordination Service Provider". The wallet still functions perfectly well as an ordinary wallet.

JoinMarket. Still going, and structurally different in a way that now looks prescient. There is no coordinator to shut down. It is a market: some participants offer liquidity for a fee, others pay it, and the matching happens peer to peer. Release 0.9.12 landed on 21 April 2026, with commits into late April. It is command-line heavy and genuinely harder to use than Wasabi, which is the price of that structure.

Samourai Wallet and Whirlpool. Gone, and not coming back. The servers and domain were seized in April 2024. Both founders pleaded guilty on 30 July 2025 to one count of conspiracy to operate a money transmitting business that transmitted crime proceeds. Keonne Rodriguez was sentenced to five years on 6 November 2025, William Lonergan Hill to four years on 19 November 2025. Any guide still telling you to open Whirlpool is describing a service that does not exist.

Status of Bitcoin privacy tools as verified on 29 July 2026
ToolStatus, 29 July 2026ModelWhat we checked
Wasabi WalletSoftware maintained, no default coordinatorCoordinated, WabiSabiv2.8.1 published 22 Jul 2026; site directs users to third-party coordinators
JoinMarketMaintained and runningPeer-to-peer market, no coordinatorv0.9.12 released 21 Apr 2026, commits to 27 Apr 2026
Samourai / WhirlpoolShut downCoordinatedServers seized Apr 2024; founders sentenced Nov 2025
PayjoinSpecified, BIP status DeployedTwo-party, no coordinatorBIP-78 header read direct from the BIPs repository
Silent paymentsSpecified, BIP status CompleteWallet-level, no soft fork neededBIP-352 header; Layer listed as Applications
Individual third-party coordinatorsNot verifiedCoordinatedWe could not confirm which are operating. Check before trusting one
The last row is deliberate. Several community-run coordinators have been announced since June 2024 and we were not able to verify independently which of them are live, honest, and still under their original operators on the day this was written. Treat any specific coordinator recommendation, including ours, as something to re-check yourself.Table by BTCLinks. Versions and BIP statuses read from the projects' own release feeds and the bitcoin/bips repository, verified 2026-07-29.

Using a wallet that improves your privacy is not itself a crime in the United States, and the Samourai case did not hold otherwise. What the founders pleaded guilty to was operating a money transmitting business that moved criminal proceeds. The government's evidence was not that they wrote privacy software; it was that they marketed the service to criminals. The sentencing release quotes Hill recommending Whirlpool to a darknet-forum user as "a much better option" for cleaning illicit funds, and Rodriguez describing the business internally as "money laundering for bitcoin".

That distinction is doing real work, and it cuts both ways. It means an ordinary person using an ordinary privacy tool is in a different position from somebody running a service. It also means the operators of coordination services now face a risk they did not price in before 2024, which is exactly why zkSNACKs stopped and why the remaining options skew towards designs with nobody at the centre to prosecute.

The practical friction is separate from the legal question. Some exchanges have flagged, delayed, or frozen deposits with recent CoinJoin history. That is a business policy, not a law, and it varies by company and by jurisdiction. If your coins are going to end up at a regulated exchange, find out that company's policy before you plan around it, not after. Nothing here is legal advice, and privacy law is very much local: this is one of the few Bitcoin topics where an hour with a lawyer in your own country is worth more than any article.

What else is there besides CoinJoin?

Quite a lot, and some of it is better suited to ordinary use.

Comparison panel. CoinJoin breaks which output is whose, needs a coordinator and other users, and is obviously detectable because of its equal-valued outputs. Payjoin breaks the assumption that inputs share an owner, needs only a willing merchant, is not detectable, and makes the recorded payment amount wrong.
Payjoin's advantage is that it hides in plain sight. Its disadvantage is that it takes two willing parties, and merchant support has been the bottleneck for years.Original diagram by BTCLinks. Mechanism per BIP-78, verified 2026-07-29.

Payjoin is specified in BIP-78, whose status in the BIPs repository reads Deployed. The merchant contributes one of their own inputs to your payment. The transaction that results looks like an ordinary two-input spend, so unlike a CoinJoin it raises no flag, and the amount recorded on chain is not the amount you paid. Every Payjoin in the wild also makes the common-input assumption slightly less trustworthy for everybody, including people who never use it.

Silent payments, BIP-352, solve a different problem: publishing a single reusable address without the address reuse that normally comes with it. The payer derives a fresh output for you from your published static address, so nobody watching the chain can total your receipts. Its BIP status is Complete and its layer is Applications, which is the detail worth noticing. It needs no soft fork, only wallets that implement it.

Plain hygiene is unglamorous and beats all of the above for most people. Never reuse an address. Do not consolidate outputs from unrelated sources into one transaction without deciding to. Keep coins from different origins in separate accounts, and label them, which is what coin control in a decent wallet is for. Our guide to address types covers why your wallet hands you a new address every time, and the threat model guide is the right place to start if you have not decided who you are actually hiding from.

Who are you doing this for?

Answer that before touching any of the tools, because the honest answers point in very different directions.

If the concern is a nosy counterparty, a landlord, an employer, or a merchant who can see your entire net worth the moment you pay them from a reused address, then address hygiene and a fresh address per payment solve most of it for free. That describes the overwhelming majority of people who look into this.

If the concern is a commercial rival reading your supplier payments off a public ledger, or a business that would rather not publish its cash position, then you are in the territory where deliberate structuring of transactions matters, and so does keeping business coins away from personal ones.

Privacy measures matched to who you are actually trying to keep information from
Who can see your financesHow commonWhat actually helps
A merchant or counterparty you paidNearly everyoneA fresh address per payment. Nothing else needed
Anyone who has your addressNearly everyoneStop reusing addresses. This is the whole fix
Your wallet's server operatorMost peopleRun your own node
A commercial rival reading the ledgerBusinessesSeparate accounts, coin control, Payjoin
A chain-analysis firm on contractRarerStructured spending over time, not one transaction
A state actor with subpoena powerRareNo consumer tool is sufficient by itself
Most people reading about CoinJoin are in the top two rows and can solve their entire problem for free. Buying tools for the bottom row while reusing addresses in the top row is the common and expensive mistake.Table by BTCLinks.

If the concern is a state actor with subpoena power, no consumer tool on this page is going to be sufficient on its own, and the failure will probably not be cryptographic. It will be the exchange that has your passport, the phone in your pocket, or the transaction you made in a hurry three years ago. Privacy is a property of a whole pattern of behaviour over years, not a button.

Bitcoin privacy leaks ranked by how often they matter and how hard they are to fix
LeakHow commonCost to fixFix
Address reuseVery commonFreeA fresh address per payment. Your wallet already does this
Querying someone else's nodeDefault behaviourOne-off hardwareRun your own node
Careless consolidationCommonFree, needs attentionCoin control; keep origins in separate accounts
Change output linkageEvery transactionAwkwardPayjoin, or careful handling of post-mix change
Amount correlationEvery transactionAwkwardEqual-output CoinJoin, splitting across spends
KYC records at an exchangeMost holdersCannot be undoneNothing retroactive. Only future choices
The top three rows account for most real-world deanonymisation and none of them need a CoinJoin. The bottom row is the one nobody can fix, which is worth knowing before spending money on the rows above it.Table by BTCLinks, ordered using the heuristics documented on the Bitcoin wiki, verified 2026-07-29.

And there is one thing worth saying plainly, because the subject attracts a lot of nervous hedging: wanting financial privacy is not suspicious. Nobody thinks it is strange that your bank statement is not published. Bitcoin is unusual among payment systems in that its ledger is public by design, so the privacy that is ambient elsewhere has to be reconstructed deliberately here. That is a property of the tool, not a confession about the user.

What actually moves the needle

In rough order of return on effort, and none of the top three involve a CoinJoin at all:

  • Stop reusing addresses. One address per payment. It is free, every wallet does it automatically, and it removes the single most reliable signal analysts have.
  • Run your own node. Otherwise you are telling somebody else's server exactly which addresses to associate with your IP, every time your wallet syncs. See running a full node.
  • Use coin control and keep origins separate. Coins from an exchange that knows your name and coins from a private sale should not meet inside one transaction unless you chose that.
  • Prefer Payjoin where a merchant offers it. No extra transaction, no fingerprint, and it helps other people too.
  • If you use a CoinJoin, plan the exit before the entry. Decide in advance how the outputs get spent, separately and unhurriedly. The mix is the cheap part.
  • Watch what a block explorer shows about your own addresses. It is the same view an adversary has, and it is free.

One last thing that should temper any confidence here. Chain analysis is an active industry with real budgets, and the heuristics in this article are the published ones. The unpublished ones are, by definition, not in any guide. Anyone who tells you a specific technique makes you untraceable is selling something, and the honest version of this subject is a running estimate rather than a solved problem. Treat privacy the way you treat security: as a habit that has to hold up over years, where one careless transaction undoes a lot of careful ones.

Sources

Some links are affiliate links. As an Amazon Associate we earn from qualifying purchases. Not financial advice.