Research Article

LiDAR: A Light-Weight Deep Learning-Based Malware Classifier for Edge Devices

Table 3

Malicious and benign features discovered by Grad-CAM.
(a)

SMS spam datasetE-mail spam dataset
No.Weight valueFeaturesNo.Weight valueFeatures

10.0023call10.0159click
20.0016free20.0141run
30.0016www30.0088could
40.0014stop40.0086file
50.0013txt50.0074remov
60.0013repli60.0070modem
70.0010cash70.0068send
800-0.0017see31,860-0.0144link
801-0.0017heart31,861-0.0178make
802-0.0018give31,862-0.0237one
803-0.0018weekend31,863-0.0334nbsp
804-0.0031get31,864-0.0454emailaddr
805-0.0038got31,865-0.1133httpaddr

(b)

Android malware dataset
20192020
No.Weight valueFeaturesNo.Weight valueFeatures

10.0118android.app->
android.view
10.0180android.app->
android.view
20.0094android.content->
android.content
20.0142android.view->
android.content
30.0088android.content->
android.app
30.0135android.os->
java.lang
40.0082android.webkit->
java.lang
40.0130android.content->
java.lang
50.0081android.app->
android.content
50.0124android.content->
android.app
60.0068android.app->android.os60.0094android.view->
android.view
70.0058android.view->
android.view
70.0092android.net->
android.net
4,667-0.0051java.net->java.lang13,201-0.0018android.content.res->
java.lang
4,668-0.0056android.view->java.lang13,202-0.0023android.database.sqlite->
android.database.sqlite
4,669-0.0064java.io->java.io13,203-0.0025android.webkit->
android.util
4,670-0.0065android.widget->
java.lang
13,204-0.0064java.io->java.io
4,671-0.0069android.content->
android.os
13,205-0.0065android.view->
android.util
4,672-0.0081android.widget->
android.util
13,206-0.0068android.widget->
android.util