Table 3: The comparison between our study and other deep learning methods.

MethodFeaturesMachine Learning ModelPrecisionRecallF1-Score

William et al. [12]API callsStacked AutoEncoders0.9550.9580.956

Zhenlong et al. [44]Permissions, Sensitive APIs, App actionsDeep Belief Networks0.9680.9680.968

Toshiki et al. [45]Malware communicationRecursive Neural Network0.9760.9620.969

Proposed SAE-DTBehavior graphsStacked AutoEncoders0.9860.9920.989