Machine-Learning Benchmarks for Phishing-and-Ransomware Detection: A Comprehensive Study

In this study, the data-driven approaches to identify and counteract the phishing and ransomware threats with the help of the machine learning tools are considered. Our assessment of four methods – Logistic Regression, Random Forest, SVM, and KNN – showed that these algorithms are efficient in the detection of cyber threats. From our experiments, Random Forest was found to have the best performance measurements with accuracy: 90%, precision: 88%, recall: 92% and F1 score of 90%. Logistic Regression next with 89% accuracy 85% precision, 91% recall values and an F1 score of 88. For SVM, the measure of accuracy was 87% while the measure of precision was 84%, recall 89% and F1 measure 86%. The “KNN” or “classical KNN” approach, although performing well, was slightly slower, with an accuracy of 81%, precision of 78%, recall of 82%, and F1 score of 80%. These results highlight the efficacy of the Random Forest model in dealing with large and intricate datasets that are associated with cyber threats. This is further underlined by the fact that it is necessary to integrate the applied technical solutions with non-functional aspects, including cybersecurity knowledge and training. The results of this study have interesting implications for developing better cyber security strategies.