Detecting malicious accounts in permissionless blockchains using temporal graph properties

Applied Network Science

Table 2 Balanced accuracy, Precision, and Recall, and F1 score for both malicious (Mal) and benign (Ben) accounts with best identified ML algorithm for supervised case when using different dataset configurations

Features	Data segment	Best classifier	Accuracy balanced	Precision		Recall		F1 score		MCC score
Features	Data segment	Best classifier	Accuracy balanced	Mal	Ben	Mal	Ben	Mal	Ben	MCC score
28	Only EOA	ExtraTrees	0.872	0.38	1.00	0.75	0.99	0.50	1.00	0.510
(PCA)	EOA and SC		0.873	0.22	1.00	0.76	0.99	0.34	0.99	0.421
36	Only EOA		0.876	0.11	1.00	0.78	0.97	0.19	0.99	0.254
36	EOA and SC		0.882	0.24	1.00	0.78	0.99	0.37	0.99	0.429
59	Only EOA		0.881	0.26	1.00	0.77	0.99	0.38	0.99	0.425
59	EOA and SC		0.887	0.29	1.00	0.78	0.99	0.42	1.00	0.473

Here, we also report the Matthews Correlation Coefficient (MCC) score
28 (PCA) EOA ExtraTreesClassifier(class_weight = ‘balanced’, max_features = 0.4, max_samples = 0.3, min_samples_leaf = 11, min_samples_split = 19, n_estimators = 600)
28 (PCA) EOA and SC ExtraTreesClassifier(class_weight = ‘balanced’, criterion = ’entropy’, max_features = 0.25, max_samples = 0.15, min_samples_leaf = 13, min_samples_split = 4, n_estimators = 800, n_jobs = 20, random_state = 100)
36 EOA ExtraTreesClassifier(bootstrap = true, class_weight = ‘balanced’, max_features = 0.15, max_samples = 0.7, min_samples_leaf = 8, min_samples_split = 18, n_estimators = 200, n_jobs = 10, random_state = 100)
36 EOA and SC ExtraTreesClassifier(class_weight = ‘balanced’, criterion = ’entropy’, max_features = 0.45, max_samples = 0.75, min_samples_leaf = 18, min_samples_split = 6, n_estimators = 200)
59 EOA ExtraTreesClassifier(class_weight = ‘balanced’, max_features = 0.2, max_samples = 0.75, min_samples_leaf = 13, min_samples_split = 19)
59 EOA and SC ExtraTreesClassifier(class_weight = ‘balanced’, criterion = ’entropy’, max_features = 0.3, max_samples = 0.3, min_samples_leaf = 14, min_samples_split = 20, n_estimators = 200)