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

Table 3

Dataset samples split 30% and 70% for test and training, respectively, for scenario 1 and scenario 2. 1999 samples from 21 families are excluded from training set for simulating zero-day attack.

		Scenario 1	Scenario 2
Training set	Malware	16982	15593
	Benign	11641	11641
Test set	Malware	7322	6712
	Benign	4988	4988
Zero-day	Malware	—	1999