Optimal Machine Learning Model to Detect Ransomware on Processor and Disk

This research study presents a machine learning-based technique for detecting ransomware, which disables devices, encrypts data, and eludes antivirus software. Because they need continuous monitoring of system messages and activity, conventional methods of detection have a significant overhead. Furthermore, these monitoring systems may be disrupted by novel ransomware variants, making identification more challenging. Our approach uses a machine learning algorithm to examine the host's CPU and disk I/O activities while ransomware is operating in a virtual machine (VM). The Harvard Dataverse V1 dataset was used to validate the method's remarkable 98.66% accuracy in identifying malware across 22 distinct ransomware occurrences and a range of user load scenarios. The 2D Convolutional Neural Network (CNN 2D) performed the best among the five methods that were studied. A comparative study also conducted with state-of-art, which is also certifying that the proposed model with 2D CNN is outperforming.