This repo contains the codes for the project detecting mixing services in bitcoin transactions via deep learning.
Keywords: graph node classification, node embedding, dynamic network, temporal random walk, skip-gram, neural network
[Wu et al.'20] Detecting Mixing Services via Mining Bitcoin Transaction Network with Hybrid Motifs, IEEE Trans. on Systems, Man, and Cybernetics, 2020. Dataset avaliable at http://xblock.pro/bitcoin/#BMD. Datasets and the core idea of the project are referred to, but we do NOT adopt the methodology in this paper.
We build a temporal network G(V,E) where V={addrs} and {weight, time} is the attribute of each edge in E. Node features are extracted via a time-respecting adaptation of node2vec. Refer to [Wen et al.'20] Bitcoin Transaction Forecasting with Deep Network Representation Learning, though the details may vary.
Then the features are fed into classifiers for classification task. We consider using LR, SVM, DNN, or LSTM.