analiza aiqbench
This study establishes an empirical baseline showing that even large 236B-parameter open-source LLMs struggle to accurately classify complex Cyber Threat Intelligence reports against the MITRE ATT&CK framework, achieving only a micro-averaged F1 score of 0.22 on unstructured data. The research highlights that model parameter size correlates with performance while prompt engineering offers negligible gains in this specific security application domain. These findings are critical for researchers aiming to build reliable AI-driven threat detection systems using open-source models rather than relying on inflated benchmark results from simplified datasets.
Evaluating Open-Source LLMs for Multi-Label ATT&CK Technique Classification on CTI Reports
arXiv:2606.18166v1 Announce Type: cross Abstract: Classifying Cyber Threat Intelligence (CTI) using MITRE Adversarial Tactics, Techniques, and Common Knowledge (ATT&CK) is essential for proactive defense, but historically required extensive human effort. Pre-Large Language Model (LLM) automation sped up this process, but could not resolve the complex language and multi-step attack patterns found in unstructured CTI reports. LLMs addressed previous limitations by using contextual reasoning to understand unstructured text. However, current evaluations rely on simplified, single-technique sentences that ignore the complexity of real-world CTI reports, which often leads to inflated performance results. Consequently, the baseline performance of open-source LLMs on complex unstructured CTI reports remains unevaluated. To address this gap, we constructed a ground-truth dataset of 2,076 human-annotated sentences (1,281 technique-positive, 795 negative) from 83 complex unstructured CTI reports. These sentences were mapped to 114 unique ATT&CK techniques using a six-phase annotation process, achieving \k{appa} = 0.68 inter-annotator agreement. Using this dataset, we evaluated seven open-source LLMs ranging from 8B to 236B parameters across prompt strategy and temperature configurations. The highest-performing LLM achieved a micro-averaged F1 score of 0.22, establishing the empirical baseline for multi-label ATT&CK classification on complex unstructured CTI. Parameter size showed a statistically significant positive correlation with F1 score. Prompt strategy and temperature produced no statistically significant gains across model configurations. These results indicate that current open-source LLMs are insufficient for production-grade ATT&CK classification. The dataset, benchmark, and findings provide a reproducible foundation for future CTI research.
Abstract:Classifying Cyber Threat Intelligence (CTI) using MITRE Adversarial Tactics, Techniques, and Common Knowledge (ATT&CK) is essential for proactive defense, but historically required extensive human effort. Pre-Large Language Model (LLM) automation sped up this process, but could not resolve the complex language and multi-step attack patterns found in unstructured CTI reports. LLMs addressed previous limitations by using contextual reasoning to understand unstructured text. However, current evaluations rely on simplified, single-technique sentences that ignore the complexity of real-world CTI reports, which often leads to inflated performance results. Consequently, the baseline performance of open-source LLMs on complex unstructured CTI reports remains unevaluated. To address this gap, we constructed a ground-truth dataset of 2,076 human-annotated sentences (1,281 technique-positive, 795 negative) from 83 complex unstructured CTI reports. These sentences were mapped to 114 unique ATT&CK techniques using a six-phase annotation process, achieving \k{appa} = 0.68 inter-annotator agreement. Using this dataset, we evaluated seven open-source LLMs ranging from 8B to 236B parameters across prompt strategy and temperature configurations. The highest-performing LLM achieved a micro-averaged F1 score of 0.22, establishing the empirical baseline for multi-label ATT&CK classification on complex unstructured CTI. Parameter size showed a statistically significant positive correlation with F1 score. Prompt strategy and temperature produced no statistically significant gains across model configurations. These results indicate that current open-source LLMs are insufficient for production-grade ATT&CK classification. The dataset, benchmark, and findings provide a reproducible foundation for future CTI research.
Cryptography and Security (cs.CR); Machine Learning (cs.LG)
arXiv:2606.18166 [cs.CR]
(or arXiv:2606.18166v1 [cs.CR] for this version)
https://doi.org/10.48550/arXiv.2606.18166 Focus to learn more arXiv-issued DOI via DataCite (pending registration)
arXivLabs is a framework that allows collaborators to develop and share new arXiv features directly on our website.
Both individuals and organizations that work with arXivLabs have embraced and accepted our values of openness, community, excellence, and user data privacy. arXiv is committed to these values and only works with partners that adhere to them.
Have an idea for a project that will add value for arXiv's community? Learn more about arXivLabs.
Text integral sub licența deschisă a publicației · 331 cuvinte
Subiecte conexe
Evaluation Sovereignty in Metadata-Driven Classification: A Multi-Track Framework for Weakly Supervised Information Systems
arXiv · 12 iun. 2026
Revisiting Accent Classification with DistilHuBERT: Speaker-Disjoint Evaluation and Cross-Corpus Generalization
arXiv · 12 oct. 2026
Dimensionality Reduction for Cyberattack Classification: A Comparative Evaluation of PCA and Linear Predictive Coding
arXiv · 6 iun. 2026
Multimodal Safety Evaluation Should Measure Controllability Beyond Classification
arXiv · 7 oct. 2026