Ransomware Detection Model Using Deep Learning Algorithms

Ransomware has emerged as one of the most destructive types of cyberattacks, harming people’s and companies’ reputations, disrupting operations, and resulting in significant financial losses. In order to increase the classification accuracy and resilience, this study created an improved ransomware detection model by combining two deep learning models with a meta-algorithm. Two benchmark datasets, the Windows PE File Analysis Dataset and the Resilient Information Systems Security Group (RISSG) ransomware dataset, were employed. After that, an autoencoder was used to optimize the features. Artificial Neural Networks (ANN) and Recurrent Neural Networks were used to classify the optimized features. The model was evaluated with performance metrics, including accuracy, precision, recall, F1-score, and time-taken. The experimental results for the two datasets showed that the best accuracy of 94.05% was obtained in RNN, the highest precision value of 0.9200 was obtained in RNN, the highest recall value of 0.9402, the highest F1-score value of 0.7855; all these best valuation metrics were recorded in RNN for the PE File Analysis Dataset. The lowest time of 0.11 was obtained in RNN for the Resilient Information Systems Security Group Dataset (RISSG Group rissggrouphubransomware dataset), when auto-encoder feature selection was used.  The findings demonstrated the ransomware detection model’s capacity to successfully identify complicated ransomware variants by achieving the highest accuracy and lowering misclassification rates when compared to traditional detection techniques. The work advanced the field of cybersecurity by introducing a scalable and intelligent ransomware detection model that integrated boosting, feature, and deep learning techniques.