Vijayakumaran, S. 2023, Analysis of cryptonote transaction graphs using the dulmage-mendelsohn decomposition. Paper presented at 5th Conference on Advances in Financial Technologies (AFT 2023).
Added by: Rucknium (2022-02-24 21:35) Last edited by: Rucknium (2024-01-10 16:22)
|Resource type: Proceedings Article
ID no. (ISBN etc.): 978-3-95977-303-4
BibTeX citation key: Vijayakumaran2021
View all bibliographic details
Creators: Bonneau, Vijayakumaran, Weinberg
Collection: 5th Conference on Advances in Financial Technologies (AFT 2023)
|Attachments 2021-760-published-version.pdf [18/37], 2021-760.pdf [20/846]
|URLs AFT conference, IACR page
CryptoNote blockchains like Monero represent the largest public deployments of linkable ring signatures. Beginning with the work of Kumar et al. (ESORICS 2017) and Möser et al. (PoPETs 2018), several techniques have been proposed to trace CryptoNote transactions, i.e. identify the actual signing key, by using the transaction history. Yu et al. (FC 2019) introduced the closed set attack for undeniable traceability and proved that it is optimal by showing that it has the same performance as the brute-force attack. However, they could only implement an approximation of the closed set attack due to its exponential time complexity. In this paper, we show that the Dulmage-Mendelsohn (DM) decomposition of bipartite graphs gives a polynomial-time implementation of the closed set attack. Our contribution includes open source implementations of the DM decomposition and the clustering algorithm (the approximation to the closed set attack proposed by Yu et al). Using these implementations, we evaluate the empirical performance of these methods on the Monero dataset in two ways -- firstly using data only from the main Monero chain and secondly using data from four hard forks of Monero in addition to the main Monero chain. We have released the scripts used to perform the empirical analysis along with step-by-step instructions.
Software with documentation: https://www.respectedsir.com/cna/
Abstract of previous version draft:
Transactions in CryptoNote blockchains use linkable ring signatures to prevent double spending. Each transaction ring is associated with a key image, which is a collision-resistant one-way function of the spent output's secret key. Several techniques have been proposed to trace CryptoNote transactions, i.e. identify the actual output associated with a key image, by using the transaction history. In this paper, we show that the Dulmage-Mendelsohn (DM) decomposition of bipartite graphs can be used to trace CryptoNote transactions. The DM decomposition technique is optimal in the sense that it eliminates every output-key image association which is incompatible with the transaction history.