Overview and Case Study for Ransomware Classification Using Deep Neural Network

Ransomware has become a prominent threat that attracted increasing attention over years. In this paper, we present a thorough overview of this particular type of malware including its initial communication channels, behavior on different platforms, infection vectors, and detection techniques. One of the problems faced research on extensive studies to detect ransomware was the unavailability of large representative datasets. In this work, we used one of the recent datasets that became publicly available with huge number of features. Here, dimensionality reduction represents a great challenge to select most relevant features. We explored different avenues of feature selection and their corresponding statistical tools in order to determine a reduced subset of features out of the entire feature haystack, using machine learning tools in python. Moreover, we ran experiments to compare a deep neural network (DNN) based mode with a random forest model for classifying ransomware. DNN showed better performance using the reduced feature set.