PDAGs & coinswaps

[nothingmuch]

We discussed the following papers:

• https://eprint.iacr.org/2023/1902.pdf
• https://informationsecurity.uibk.ac.at/pdfs/MB2017_AnonymousAlone.pdf

The first introduces a general privacy DAG (PDAG) model for blockchains, that is able to model privacy preserving cryptocurrencies like monero or zcash with the inclusion of more kinds of edges than the Bitcoin DAG's spends.

In the context of Bitcoin, this can be used to represent potentially related coins due to graph disjoint privacy techniques, such as CoinSwap. An early attempt at quantifying an upper bound for CoinSwap anonymity sets was provided in the 2nd paper, we discussed how CoinSwap has evolved since then and how it would look like if instantiated as a PDAG.

More generally we discussed how all clustering heuristic results can as labeled edge sets on the transaction graph, how the previously discussed results of Moser & Narayanan can be reframed in this perspectivee (random forest or other decision tree based machine learning techniques such as extreme gradient boosting), the tension between local judgements and the transitivity property of equivalence relations that makes clustering a global optimization problem, and how this relates to modular decomposition of graphs.