Combat Mobile Evasive Malware via Skip-Gram-Based Malware Detection

Table 4

Scenario 1 test results for different machine learning methods. Ensemble 1 is majority voting of all other methods shown in this table. Random forest bested other methods with 95.64% accuracy.

	Malware			Benign			Total accuracy
	Precision	Recall	F1 measure	Precision	Recall	F1 measure
SVM	86.5	79.7	82.9	73.2	81.7	77.2	80.48%
Random forest	97	95.6	96.3	93.7	95.7	94.7	95.64%
Decision tree	91.9	93.6	92.7	90.3	87.9	89.1	91.25%
Random subspace	95.7	94.4	95.1	92	94.8	92.9	94.18%
SGD	82.9	93.4	87.8	88.8	71.7	79	84.58%
KNN	93.4	96.9	95.1	95.2	90	92.5	94.09%
Ensemble 1	95	96.5	95.7	94.7	92.5	93.7	94.88%