Cyber Threat Intelligence Extraction Using Text Mining Techniques for Proactive Security

Cyber Threat Intelligence (CTI) is valuable in gaining insight into current tactics, techniques and procedures used by the adversaries in a cyber attack. In this work, a new framework applying GAN for classifying tactics and techniques from unstructured CTI data is presented. The described approach combines two models by addressing the link between high-level tactics and low-level middle-level technologies by using a hierarchical method and CNNs based on attention mechanisms. Encoders from the specific subject fields as well as high-quality preprocessing methods are used to transform the initial information from the CTI reports into the set of features which could be easily utilized. Thus, the adversarial training mechanism ensures that the generated data are reasonable and semantically correct to promote enhanced classification. The proposed framework was tested on several datasets, and the experimental results demonstrated higher accuracy, precision, recall, F1 score than other relevant approaches. The imposed hierarchy on GAN improves system's realism in feature representation, supporting CTI analysis and cyberattack behavior comprehension. This approach establishes the groundwork for the implementation of large-scale, preventive attitude into the cybersecurity models, and arms analysts with information obtained from difficult and chaotic data feedstocks.