AI-Driven Proactive Framework for Cybersecurity Threat Prediction, Detection, and Attack Classification
Abdel Rahman Alkharabsheh, Fatima Hassan Alhosani, Mariam Hasan Alameri, Alyazia Bakhit Alrashdi, Fatima Mubarak Almenhali, Amna Aref Alzaabi · 2025
In a rapidly-evolving digital world businesses face more aggressive cyber threats and the challenges of cybersecurity. Conventional systems, that rely on reactive solutions (like firewall and signature-based intrusion detection systems (IDS)), cannot protect against modern attacks. These traditional methods only trigger after a breach has happened, which usually leads to the loss of sensitive data and has both operational and financial implications. This reactive orientation highlights a fundamental shortcoming of most extant research; namely, that there currently exists no proactive, intelligent system that enables an organization to predict and prevent threats before they appear. To fill this void, this paper brings a new AI-centered cybersecurity framework, pivoting on defense strategy from reactive to proactive. Using models such as Logistic Regression, Random Forests, and custom risk scoring functions, the system is designed to identify, classify, forecast cyber threats in real time. Unlike static models, this system leverages external threat intelligence, behavior modeling, and anomaly detection to continuously assess risk and apply appropriate countermeasures. Its comprehensive and flexible framework of machine learning (including predictive analytics, auto classification, and meaning-based AI), ensures that the platform is versatile, scalable and transparent- characteristics which are critical in modern enterprise environments. The system was emulated with frameworks such as Scikitlearn and TensorFlow and was effective in several evaluation metrics: accuracy over 91 %, precision 92.1 % and F1 90.9 %. These findings confirm that the proposed model can effectively identify the malicious activities while it delivers a low false positive and false negative rate. The incorporation of interpretable features like SHAP promotes trust and usability with cyber security professionals. Such a framework is not only increasing detection rates but also enabling intelligent, instantaneous decisions that cut the risk to an organization and increase compliance to mandates. This work is a significant advance in the field of cybersecurity, building an intelligent and proactive defense system that can evolve according to a dynamic cyber-threat environment. The resulting system is not only scalable and interpretable, it also has a high degree of accuracy, and therefore is highly applicable in practical scenarios. This research benefits the development of cutting-edge security measure against possible vulnerabilities, tackling the evolving cyber threats.