With the widespread use of mobile devices, Android has
become an important target for malware attacks because of its
open ecosystem and flexible permission mechanism. Based on
this background, this thesis designs and implements a deep
learning-based A...
With the widespread use of mobile devices, Android has
become an important target for malware attacks because of its
open ecosystem and flexible permission mechanism. Based on
this background, this thesis designs and implements a deep
learning-based Android malware detection system. The system
uses fully connected layers and a lightweight attention
mechanism to process 470-dimensional behavior features from
Android samples. It also provides a Flask-based web interface,
through which users can upload CSV feature files, select a
detection model, and check the classification result.
Experiments on the CICMalDroid2020 dataset show that the
proposed model achieved 0.92 accuracy, with precision, recall,
and F1-score values of 0.90, confirming its usability in the
tested five-class classification task.