NLP Technique for Malware Detection Using 1D CNN Fusion Model

Table 6

Comparative analysis with peer technique models.

	Architecture	Dataset (no.)	Features	Feature representation	Results (%)
Jung et al. [35]	Inception-v3 and Inception-ResNet-v2 (2D CNN)	Malware: 5377; Benign: 6249	Data section bytes	Grayscale image	Acc: 98.02
Aksakalli [34]	2D CNN	Malware: 2500; Benign: 2500	Permissions	One-hot sparse vector	Acc: 96.71
Hasegawa and Iyatomi [36]	1D CNN	Malware: 5000; Benign: 2000	Raw APK bytes	Byte representation	Acc: 97.04
Our model	Ensemble 1D CNN	Malware: 4948; Benign: 2477	n-gram static opcodes	Word2vec embedding	Acc: 95.5; Precc: 98.0