This post explores using sequence modeling, specifically LSTM and BERT, to detect Android malware in highly imbalanced datasets. The application of BERT achieves excellent results with an F1 score of 0.919 on imbalanced malware detection.
Table of contents
A Summary of What I Have Learned!Main ReferenceThe Challenge of Imbalanced Malware DatasetsSequence Modeling Retains Ordering InformationApplying BERT Language Model to Security SequencesOrdering of Activities is CriticalPotential for Security Sequence AnalysisDiscussion Takeaways10 Impressions