<![CDATA[This article focuses on using deep learning methods to predict ransomware from hybrid analysis samples. Other similar research is analysed to understand the common methods of detection used to predict ransomware using various methods of analysis. By using this knowledge an experiment is created which tests the performance of a model created from hybrid analysis of ransomware samples. The training dataset used is made up of more than five hundred samples containing 38 different ransomware families and benign Windows program samples. The resultant model was then tested against a dataset include ransomware families not represented in the training dataset, which showed a decrease in performance. These results were then compared to other research’s reported results which highlights potential issues in the way artificial intelligence models are tested and reported. The paper then proposes a focus on more complex methods of prediction, and other potential methods to ensure the models created are externally as effective as they report.]]>
Copyrights © 2026
