Malware is constantly evolving in a fast-paced technological society and now encompasses a wide range of attack vectors that were not a concern a decade ago. As a result, in order to keep in pace with the advancement of technology, it is imperative that analysis and detection of malware follows the same trend. The purpose of this study is to discover if there are any underlying relationships or trends between different malware families.
- Malware samples collected from Cuckoo Web Interface
- Static and Dynamic features extracted
- Static, Dynamic and Hybrid dataset constructed
- Supervised Leanring models(Logistic Regression, Random Forest, XGBoost, Neural Networks) applied on all 3 datasets to build and train the classifiers
- Grid search for optimal parameters
Best Performer: Neural Network with 2 hidden layers
Best Approach: Hybrid Analysis
Expand dataset: May help improve performance values
Dr. Matthew Elder, Johns Hopkins University Applied Physics Laboratory
William J. La Cholter, Johns Hopkins University Applied Physics Laboratory
Dr. Xiangyang Li, Johns Hopkins University