Malware Detection on Byte Streams of PDF Files Using Convolutional Neural Networks

Table 3

Experimental results with the PDF dataset, where the two values of each cell are of ‘benign’ and ‘malicious’, respectively.

Model	Precision	Recall	F1
DT	96.00 / 90.30	89.70 / 96.30	92.70 / 93.20
NB	88.40 / 99.70	99.70 / 87.00	93.70 / 92.90
SVM	94.70 / 98.90	99.00 / 94.40	96.80 / 96.60
RF	93.50 / 99.40	99.40 / 93.10	96.40 / 96.10
Emb+Conv+Conv+Pool+FC	99.76 / 100.0	97.37 / 97.37	98.48 / 98.65
Conv+Conv+Pool+FC	99.78 / 100.0	92.62 / 97.27	95.71 / 98.61
Emb+Conv+Pool+FC	99.73 / 100.0	94.94 / 97.78	97.12 / 98.87
Emb+Conv+Conv+FC	99.67 / 100.0	97.78 / 92.32	98.55 / 96.00
Emb+Conv+Conv+Conv+Pool+FC	99.70 / 100.0	92.21 / 95.35	95.36 / 97.60