Image Classification-Based Ransomware Attack Detection Using Deep Learning Algorithm

Ransomware represents a significant cybersecurity threat that can potentially damage corporate entities. Conventional antivirus methodologies must be revised to mitigate emerging and complex cyber-attack vectors. This study introduces an architecture for a Convolutional Neural Network (CNN) predicated on deep learning paradigms; it utilizes a dataset composed of both malicious and benign executables. These files were transformed into grayscale images via the Python Imaging Library (PIL) and subsequently processed by the CNN to effectively discriminate between images indicative of ransomware and those classified as benign. The effectiveness of the CNN model was measured using metrics like accuracy, precision, recall, and the F1 score. The training accuracy was 0.9476, while the validation accuracy was 0.9383 percent. Moreover, the precision, F1 score, and recall of training were higher than their respective validation rates. This model can contribute to developing an effective solution for detecting ransomware attacks, complement existing approaches, and improve accuracy.