Skip to main content

Table 1 Experimental results on Kaggle data. In Tables 1 and 2, the best experimental results are shown in boldface

From: DeepMal: maliciousness-Preserving adversarial instruction learning against static malware detection

Classifier Baseline Obfuscation-based Feature-level DeepMal
  ACC mF1 ACC mF1 ΔmF1 ACC mF1 ΔmF1 ACC mF1 ΔmF1
LR 0.9813 0.9812 0.9771 0.9769 0.0043 0.5103 0.5059 0.4753 0.8538 0.8461 0.1351
SVM 0.9903 0.9903 0.9825 0.9823 0.0080 0.6135 0.6124 0.3779 0.4589 0.3155 0.6748
RF 0.9855 0.9853 0.9783 0.8100 0.1753 0.5978 0.5912 0.3941 0.8126 0.7980 0.1873
Lenet-5 0.9728 0.9726 0.9529 0.9524 0.0202 0.2572 0.2497 0.7229 0.0332 0.0332 0.9394
All-Cov 0.9746 0.9745 0.9299 0.9244 0.0501 0.3900 0.3839 0.5906 0.0577 0.0575 0.9170
NiN 0.9487 0.9484 0.9281 0.928 0.0204 0.3804 0.3738 0.5746 0.1285 0.1285 0.8199
VGG16 0.9487 0.9484 0.9281 0.9280 0.0204 0.3804 0.3738 0.5746 0.1285 0.1285 0.8199
ResNet 0.9215 0.9213 0.8991 0.8983 0.0230 0.4215 0.4025 0.5188 0.2022 0.2018 0.7195
Standard DNN 0.9134 0.9098 0.8999 0.8901 0.0197 0.6032 0.5832 0.3266 0.3023 0.3525 0.5573