Machine Learning Based Hybrid Approach in Ransomware Recognition and Classification

Cyber security is severely restricted by spyware, ransomware, along malevolent assaults, which can seriously harm networks, server rooms, websites, and mobile devices in a variety of commercial and industrial settings. Conventional anti-ransomware software finds it difficult to defend against immediately developed, highly competent attacks. As a result, contemporary techniques such as conventional and neural network-based topologies can be greatly applied to creating novel ransomware remedies. This research work employs a feature selection-based method along with implementing machine learning classification approaches in ransomware malware recognition and classification. Moreover, we developed six machine learning approaches: Adaptive Boosting, K-Nearest Neighbor, Stochastic Gradient Descent, Extra tree, Artificial Neural Network and Hybrid approaches based on preferred features for ransomware malware classification. Our investigational outcomes reveal that the proposed hybrid model outperforms conventional approaches with a detection accuracy of 99.5% in terms of measures like accuracy, precision, F1-score, Recall, Matthew’s Correlation Coefficient and Kappa score.