Choose a system to see explanations


  • The Gallery consists of examples from the four datasets used in our paper with different neural network architectures:
    • Drebin: Android malware detection using static string features from the manifest file (MLP).
    • Mimicus: Malicious PDF file detection using static document features (MLP).
    • VulDeePecker: Detection of vulnerabilities in code snippets (LSTM).
    • DAMD: Android malware detection using decompiled Dalvik byte-code (CNN).
  • You can toggle on/off the explanation of a method by clicking on its name.
  • For the static features (Drebin, Mimicus) we offer two ways of sorting (located in top right)
    • Key: Sorts features alphabetical to compare them side by side.
    • Value: Sorts features from largest to smallest to compare most important features side by side.
  • The "max-features" field allows you to see only the N most important features.