Abstract

This academic study analyzes machine learning methods for CVE classification. Machine learning models are tested to improve cyber threat detection and mitigation. The study examined Decision Trees, Logistic Regression, Gaussian Naive Bayes, Support Vector Machines (SVM), a random forest classifier, and a simple neural network. The models were assessed using accuracy, precision, recall, F1 Score, F2 Score, AUC-ROC, and AUC-PR. The SVM and Random Forest Classifier were the most accurate before Grey Wolf Optimization. The Simple Neural Network was more accurate and recallable. The Grey Wolf Optimization approach improved the Support Vector Machine (SVM) and Random Forest Classifier models' precision to 0.97. This shows Grey Wolf Optimization could classify vulnerabilities well. The pros and cons of using machine learning methods like Grey Wolf Optimization in cybersecurity are examined in this paper. Vulnerability categorization helps identify and mitigate threats quickly. Models and optimization can help cybersecurity experts prioritize vulnerabilities and improve defenses. This research advances cybersecurity machine learning by demonstrating algorithm and optimization effectiveness. However, the publication acknowledges its shortcomings and suggests future research into dataset volume, inclusivity, and model efficacy. Academics may study multi-modal learning and real-time analytic methods to improve cybersecurity. In conclusion, improved machine learning algorithms improve vulnerability category understanding and repercussions. Comparative analysis and Grey Wolf Optimization improve cybersecurity measures in an ever-changing cyber environment, protecting important assets and data.