Aparecium: understanding and detecting scam behaviors on Ethereum via biased random walk

Cybersecurity

Table 2 Performance comparison in different models and graph generation methods

		Random time period based	Random address based	Policy-based selective addresses
		Li’s (Li et al. 2021)	Chen’s (Chen et al. 2020a)	Chen’s (Chen et al. 2021)	Wu’s (Wu et al. 2020)	Ours
Deepwalk	Precision	0.318	0.595	0.782	0.799	0.911
	Recall	0.518	0.158	0.727	0.762	0.729
	F1-score	0.394	0.250	0.753	0.780	0.810
Node2vec	Precision	0.364	0.648	0.827	0.870	0.864
	Recall	0.543	0.157	0.749	0.822	0.842
	F1-score	0.436	0.253	0.786	0.845	0.853
GCN	Precision	0.417	0.628	0.881	0.932	0.984
	Recall	0.580	0.174	0.719	0.720	0.848
	F1-score	0.485	0.272	0.792	0.813	0.911
GraphSage	Precision	0.387	0.610	0.854	0.970	0.949
	Recall	0.569	0.154	0.703	0.746	0.851
	F1-score	0.461	0.246	0.771	0.844	0.897