Understanding coinjoin transactions

Coinjoin implementations

Privacy on Bitcoin

Coinjoin implementations

JoinMarket
Wabisabi

For the moment, there are mainly 3 different coinjoin implementations on Bitcoin:

Whirlpool;
Wabisabi;
JoinMarket.

In addition to these three historical implementations, a new one has recently emerged: Joinstr. I will not address it in this chapter, but you can find all related information in this dedicated tutorial:

JoinstrDecentralized CoinJoins via the Nostr network for sovereign Bitcoin confidentiality

Each of these implementations aims to break the history of UTXOs via coinjoin transactions. However, their mechanisms vary considerably. It is therefore essential to understand how each works, so you can choose the option that best suits your needs.

JoinMarket

JoinMarket, founded in 2015 by Adam Gibson and Chris Belcher, stands out clearly from other coinjoin implementations thanks to its unique model for connecting users. The system is based on a P2P exchange market where some users, the "makers", make their bitcoins available for mixing, while others, the "takers", use this cash to make coinjoins in return for a fee.

In this model, "makers" make their bitcoins available to "takers" and receive a fee for their service. The takers, in turn, pay to use the makers' bitcoins to carry out their own coinjoin transactions. Service fees vary according to the role occupied: "makers" accumulate fees for offering liquidity, while "takers" pay the fees. The market operates freely, with no conditions of use.

One of JoinMarket's main drawbacks is its complexity of use, which requires a certain degree of comfort with terminals to operate it effectively. While this complexity is not an obstacle for the experienced user, it may limit access to the general public. However, the recent introduction of a web interface called JAM has made it a little easier to use.

Source: JAM

However, the technical barrier remains a major obstacle. In the coinjoin ecosystem, where confidentiality is reinforced by the number of participants, any limitation reducing accessibility directly affects the available liquidity, which is a crucial factor in the efficiency of the mix. Bitcoin, already a niche in financial transactions, sees its use of coinjoins as a sub-niche, and JoinMarket represents an even more specialized fraction of it, which therefore restricts its potential to increase the number of users' anonsets.

Despite its innovative P2P linking model for coinjoiners, JoinMarket has some significant disadvantages, particularly in terms of transactional structure. Unlike other implementations, such as Whirlpool, JoinMarket does not guarantee perfect equality between outputs, and it is possible to trace deterministic links between inputs and outputs. Moreover, it lacks tools to prevent UTXOs that have already been mixed together from being mixed again, which could compromise the confidentiality that users seek.

Finally, while the JoinMarket concept is interesting, especially for those interested in a dynamic liquidity market, its structural weaknesses and technical complexity make it, in my opinion, less appealing for both novices and experts seeking a coinjoin implementation.

JoinMarketGuide and tutorial on how to use JoinMarket to do CoinJoin over Bitcoin via command line

Wabisabi

Wabisabi is another coinjoin implementation, with an approach that centralizes transaction coordination. This model was conceived by Ádám Ficsór (nopara73), Yuval Kogman, Lucas Ontivero, and István András Seres in 2021, and was integrated into Wasabi 2.0 software the following year. Wabisabi is precisely an evolution of the Wasabi software coinjoin model launched in 2018.

Towards the end of the 2010s, Wasabi adopted a radically different coinjoin transaction structure to Whirlpool. Wasabi used very large coinjoin transactions involving dozens of participants to increase the anonsets of its participants. In contrast, Whirlpool opted for multiple small transactions, enabling anonsets to grow exponentially with each cycle.

Exchange management methods also distinguished the two implementations. With Whirlpool, the change was excluded and isolated from UTXOs prior to coinjoin cycles thanks to TX0, a concept I'll explain further in the next chapter. With Wasabi, on the other hand, the change formed one of the outputs of the coinjoin transaction, maintaining deterministic links between certain inputs and outputs.

With Wabisabi, version 2.0 has adapted its approach to coinjoins to move closer to that of Whirlpool. Although coinjoin transactions remain very large, it is now possible to chain several successive cycles, thus following the Whirlpool model. Particular effort has also been made regarding change management: unlike Wasabi 1.0, where change was directly linked to users’ inputs, Wabisabi seeks to subdivide change into several smaller amounts, distributed in equal denominations among all participants.

Let's illustrate this with a simplified example involving just two users: Alice wishes to mix 115,000 sats, and Bob wishes to mix 210,000 sats. Ignoring fees, with Wasabi 1.0, a coinjoin transaction would have generated 3 outputs of 100,000 sats, plus 1 change of 15,000 sats for Alice and 1 change of 10,000 sats for Bob. The change outputs would still be linked to the inputs:

Under Wabisabi, the same transaction would have produced 3 outputs of 100,000 sats and 5 outputs of 5,000 sats, thus dispersing the change so that it could not be directly linked to a specific input:

Personally, I find that Wabisabi's change management presents several risks that could compromise its effectiveness in terms of confidentiality:

When a user contributes a UTXO that is significantly larger than those of other participants, he inevitably ends up with a change amount that will be linked to his input. This runs counter to the original aim of the protocol, which is to eliminate all identifiable changes.
The multiplication of denominations with the aim of fragmenting the change can paradoxically be detrimental to mixing efficiency. This process can lead to a reduction in anonsets for certain outputs, as they become more easily identifiable;
This method also generates low-value UTXOs, which pose a management problem for the user. These small UTXOs, if they become too costly to spend in relation to their value, can become "dust". This phenomenon leads the user to merge several UTXOs into inputs for future transactions or to consolidate them into a single UTXO. In both cases, due to the CIOH, this can either reduce the anonsets obtained or completely cancel out the confidentiality benefits acquired by the initial coinjoin.

Unlike Whirlpool, which implements the ZeroLink protocol to ensure rigorous separation between pre-mix and post-mix UTXOs, Wabisabi does not maintain this strict segregation. There have also been problems of address reuse by some Wasabi customers, which are obviously very detrimental to the user.

In Wasabi version 2.0, a new coinjoin fee policy has been implemented. From now on, coordinator fees are set at 0.3% for UTXOs above 0.01 bitcoin, while for smaller UTXOs, these fees are offered at a full rate. Additionally, remixes for these smaller UTXOs are provided at no charge, although mining fees remain payable by the user for all transactions, including remixes.

This contrasts with Whirlpool's policy, where fees remain fixed, regardless of the size of the anonsets obtained. With Wasabi 2.0, although coordinator fees are waived for small UTXOs, users still incur mining fees on all transactions, including remixes.

As I write these lines, the use of Wabisabi has become significantly more complex due to recent events. Following the arrest of Samourai Wallet's founders, zkSNACKs, the company that finances and manages Wasabi's development, announced that its coinjoin coordinator service would be discontinued on June 1, 2024. This coordinator, which was set up by default on Wasabi, was responsible for the vast majority of liquidity.

With the discontinuation of this main coordinator, users must now connect to new, independent coordinators. This change raises several concerns: on the one hand, new coordinators may not have sufficient liquidity, which could reduce the effectiveness of coinjoins in terms of confidentiality. On the other hand, there is a risk of encountering a malicious coordinator. This situation adds significant new risks for those seeking to use Wabisabi.

Beyond the technical issues, the decision by zkSNACKs, the company behind Wasabi, to use the services of a string analysis company to filter coinjoin participants raises serious ethical and strategic questions. The initial idea was to prevent criminals from using coinjoins on Wasabi, a move that may seem legitimate. However, it raises a paradox: paying fees to a coordinator whose primary mission is to reinforce user confidentiality, only to have him fund a company whose aim is to compromise that same confidentiality.

Even more worrying is the principle of filtering, which contrasts radically with Bitcoin's philosophy of offering an open, uncensored financial system. While it may seem justified to exclude criminal activities, this filtering could also affect individuals whose actions, although classified as illegal in certain contexts, may be morally justifiable or socially beneficial. The example of Edward Snowden perfectly illustrates this dichotomy: considered a criminal by some governments for his revelations, he is seen by others as a whistleblower who acted in the public interest. This complexity underlines the potential danger of filtering, which, although well-intentioned, can ultimately undermine the rights and security of legitimate users. I could also have mentioned activists and journalists who are persecuted under certain authoritarian regimes.

As you'll have gathered by now, my preference is definitely for the Whirlpool model for coinjoins on Bitcoin. This system stands out for its rigor and offers superior guarantees of confidentiality. It is also the only one to offer a mix considered perfect in a mathematical context. In my opinion, this model represents the future of coinjoins on the Bitcoin network. I invite you to explore this model in greater depth in the next chapter.

Wasabi Walletlearn how to regain your privacy

Ginger WalletOpen-source, self-custody Bitcoin wallet software, fork from Wasabi Wallet, integrating Coinjoins

Quiz

btc2045.3

1/5

How are makers compensated in JoinMarket?