Explainable Artificial Intelligence-Based IoT Device Malware Detection Mechanism Using Image Visualization and Fine-Tuned CNN-Based Transfer Learning Model

Table 2

Comparison of accuracy and loss between four different CNN pretrained models.

Dataset 1
Pretrained model	Train accuracy	Train loss	Test accuracy	Test loss
DenseNET201 (TL)	0.963	0.093	0.947	0.127
MobileNET (TL)	0.964	0.092	0.942	0.195
ResNET50 (TL)	0.929	0.18	0.868	0.303
Inception-v3 (TL + FT)	0.992	0.023	0.969	0.107
Dataset 2
DenseNET201 (TL)	0.803	0.629	0.757	0.804
MobileNET (TL)	0.792	0.637	0.764	0.805
ResNET50 (TL)	0.449	1.686	0.387	1.779
Inception-v3 (TL + FT)	0.922	0.021	0.786	0.073

TL, transfer learning; FT, fine-tuning.