• @james_waugh_28 30 January 2020

  If you're familiar with Integer Programming, the algorithm I used is an integer program where the decision variables are binary (0/1) inclusion variables for each of the inputs / outputs. The objective is to maximise the size of the balanced subset (sum of the binary inclusion variables multiplied by their values). The constraint is to ensure that the sum of the inputs (times their decision variables) balances with the sum of the outputs (times their decision variables) (to within a tolerance). Then I just added a few more constraints to restrict the search space (e.g. a constraint that the size of members of the matched subset must be less than 6).

• @nyusternie 31 January 2020

  and what about tx fees? wouldn't that add to the complexity of matching inputs and outputs? > It could be that Alice took 1+3=4 satoshis, paid the miner nothing and left with 4 satoshis. But it's totally possible otherwise (which is correct) that she brought 1+3, gave miner 1 satoshi and left with 3. (In reality, of course, you must pay miners fee) (https://read.cash/@Read.Cash/cashfusion-on-bitcoin-cash-what-it-is-and-lets-try-to-crack-it-e20462a5)

  • @james_waugh_28 1 February 2020

    That's a great point. For the purposes of this article I assumed that transaction fees would be charged equally to each input contributed into the transaction, however the CashFusion devs have previously stated that they could potentially 'randomise the transaction fees to make identifying matches even more difficult', doing so would make it far more unlikely that anyone could identify a 'true match'.

• @JZA 3 January 2026

  CF should be easy to implement through programming libraries so applications can ofuscate payments by default.

• @JZA 26 August 2026

  Just wish more wallets from BCH had the protocol. But many still lack them.

More comments