Title:Linking Common Vulnerabilities and Exposures to the MITRE ATT&CK Framework: A Self-Distillation Approach

Abstract:Due to the ever-increasing threat of cyber-attacks to critical cyber infrastructure, organizations are focusing on building their cybersecurity knowledge base. A salient list of cybersecurity knowledge is the Common Vulnerabilities and Exposures (CVE) list, which details vulnerabilities found in a wide range of software and hardware. However, these vulnerabilities often do not have a mitigation strategy to prevent an attacker from exploiting them. A well-known cybersecurity risk management framework, MITRE ATT&CK, offers mitigation techniques for many malicious tactics. Despite the tremendous benefits that both CVEs and the ATT&CK framework can provide for key cybersecurity stakeholders (e.g., analysts, educators, and managers), the two entities are currently separate. We propose a model, named the CVE Transformer (CVET), to label CVEs with one of ten MITRE ATT&CK tactics. The CVET model contains a fine-tuning and self-knowledge distillation design applied to the state-of-the-art pre-trained language model RoBERTa. Empirical results on a gold-standard dataset suggest that our proposed novelties can increase model performance in F1-score. The results of this research can allow cybersecurity stakeholders to add preliminary MITRE ATT&CK information to their collected CVEs.