Attribution Classification Method of APT Malware in IoT Using Machine Learning Techniques

Table 3

Classification results of each model on each APT organization.

APT organization	Evaluation	KNN	DT	XGB	SMOTE-RF
Lazarus group	Precision	0.791	0.750	0.800	0.845
	Recall	0.507	0.493	0.478	0.567
	F-score	0.618	0.595	0.598	0.644
APT28	Precision	0.360	0.351	0.355	0.366
	Recall	0.854	0.833	0.792	0.854
	F-score	0.506	0.494	0.490	0.513
Operation C-Major	Precision	0.889	0.889	0.889	0.889
	Recall	0.828	0.828	0.828	0.828
	F-score	0.857	0.857	0.857	0.857
APT29	Precision	0.937	0.938	0.912	0.968
	Recall	0.825	0.839	0.857	0.834
	F-score	0.877	0.886	0.884	0.896
Dropping Elephant	Precision	0.927	0.980	0.944	0.927
	Recall	0.836	0.836	0.836	0.836
	F-score	0.879	0.903	0.887	0.879
Sandworm	Precision	0.840	0.917	1.0	1.0
	Recall	0.808	0.846	0.846	0.885
	F-score	0.824	0.880	0.917	0.939
Naikon	Precision	0.913	0.957	0.917	0.957
	Recall	0.700	0.733	0.733	0.733
	F-score	0.792	0.830	0.815	0.830