Proposed Ransomware Detection Model Based on Machine Learning

Abstract

Ransomware is one of the main malwares that exists today as established by EUROPOL and Malwarebytes which affects both the international and national context. In this way, the main problem is the detection of ransomware in the different users. Due to the lack of a responsive security control against the mentioned malware that adapts to the different variants that may arise, since currently virus signature is used which is not effective because of its dependence on manual updates. Thus, the overall objective is to develop the proposal of a Machine Learning logic model to improve the detection of Ransomware. For this purpose, the proposed methodology was used, due to its adaptability in predicative research. The results obtained from the model selection and training process showed that the Random Forest algorithm had the highest accuracy, and when trained by means of a Dataset for ransomware detection, 0.99 in Accuracy, 0.99 in Balanced Accuracy, 0.99 in ROC AUC and 0.99 F1 Score were obtained. Thus, it is proved that the Random Forest model is the ideal model for ransomware detection due to its effectiveness and accuracy. © The Author(s), under exclusive license to Springer Nature Switzerland AG 2024.