Since cyber-attacks are ever-increasing in number, intensity, and variety, a strong need for a global, standardized cyber-security knowledge database has emerged as a means to prevent and fight cybercrime. Attempts already exist in this regard. The Common Vulnerabilities and Exposures (CVE) list documents numerous reported software and hardware vulnerabilities, thus building a community-based dictionary of existing threats. The MITRE ATT&CK Framework describes adversary behavior and offers mitigation strategies for each reported attack pattern. While extremely powerful on their own, the tremendous extra benefit gained when linking these tools cannot be overlooked. This paper introduces a dataset of 1813 CVEs annotated with all corresponding MITRE ATT&CK techniques and proposes models to automatically link a CVE to one or more techniques based on the text description from the CVE metadata. We establish a strong baseline that considers classical machine learning models and state-of-the-art pre-trained BERT-based language models while counteracting the highly imbalanced training set with data augmentation strategies based on the TextAttack framework. We obtain promising results, as the best model achieved an F1-score of 47.84%. In addition, we perform a qualitative analysis that uses Lime explanations to point out limitations and potential inconsistencies in CVE descriptions. Our model plays a critical role in finding kill chain scenarios inside complex infrastructures and enables the prioritization of CVE patching by the threat level. We publicly release our code together with the dataset of annotated CVEs.

Effectively understanding and categorizing vulnerabilities is vital in the ever-evolving cybersecurity landscape, since only one exposure can have a devastating effect on the entire system. Given the increasingly massive number of threats and the size of modern infrastructures, the need for structured, uniform cybersecurity knowledge systems arose. To tackle this challenge, the MITRE Corporation set up two powerful sources of cyber threat and vulnerability information, namely the Common Vulnerabilities and Exposures (CVEs) list focused on identifying and fixing software vulnerabilities, and the MITRE ATT&CK Enterprise Matrix, which is a framework for defining and categorizing adversary actions and ways to defend against them. At the moment, the two are not directly linked, even if such a link would have a significant positive impact on the cybersecurity community. This study aims to automatically map CVEs to the corresponding 14 MITRE ATT&CK tactics using state-of-the-art transformer-based models. Various architectures, from encoders to generative large-scale models, are employed to tackle this multilabel classification problem. Our results are promising, with a SecRoBERTa model performing best with an F1 score of 77.81%, which is closely followed by SecBERT (78.77%), CyBERT (78.54%), and TARS (78.01%), while GPT-4 showed a weak performance in zero-shot settings (22.04%). In addition, we perform an in-depth error analysis to better understand the models’ performance and limitations. We release the code used for all experiments as open source.