The increasing complexity of modern digital infrastructures has made cybersecurity threat detection more difficult and operationally demanding. Traditional rule-based and signature-driven systems remain effective for known attack patterns, but they are often unable to detect emerging, modified, or previously unseen threats. This paper proposes a practical AI-driven framework for cybersecurity threat detection that integrates data collection, preprocessing, feature engineering, machine learning-based detection, threat classification, and response handling within a unified architecture. The framework emphasizes operational usability, scalability, and deployment relevance. Experiments conducted using NSL-KDD and UNSW-NB15 datasets demonstrate improved detection performance and reduced false positives compared with traditional approaches. The proposed framework bridges the gap between theoretical machine learning models and real-world cybersecurity implementation.
Muhammad Golam Soroar (Mon,) studied this question.