Evaluation of Deep Learning Methods Efficiency for Malicious and Benign System Calls Classification on the AWSCTD

Table 5

Malicious and benign samples (malware or clean) classification accuracy.

Count	MalwarePlusClean				MalwarePlusClean2
	LSTM	GRU	CNN	SVM	LSTM	GRU	CNN	SVM
10	87.0	87.4	87.4	87.4	88.6	88.6	88.9	88.4
20	90.0	90.2	91.1	90.2	92.0	92.4	93.0	91.8
40	91.2	90.6	91.8	90.5	92.9	92.4	94.1	92.0
60	91.2	91.1	93.3	91.1	93.2	93.0	93.5	91.9
80	92.7	92.0	94.2	91.5	93.0	93.2	94.5	91.7
100	92.7	92.5	94.3	91.5	93.0	92.1	94.5	91.2
200	90.7	90.8	94.3	88.6	92.4	91.8	94.9	88.9
400	87.0	90.9	94.8	88.4	89.2	92.0	94.8	88.7
600	86.4	88.7	98.5	87.2	85.8	90.1	98.6	87.1
800	84.2	84.7	98.8	86.9	85.1	82.8	98.9	87.0
1000	84.2	70.3	98.9	87.2	83.8	72.4	99.3	87.0