TIRE: Advancing Threat Intelligence Relation Extraction with a Novel Data-Centric Framework
Bibliographic record
Abstract
Relation extraction (RE) plays a critical role in uncovering hidden connections and assisting security analysts in identifying complex patterns within cyber threat intelligence (CTI) data. Despite its importance, RE faces significant challenges, such as overlapping relations, complex sentence structures with long-range dependencies, and ambiguous relation types. Existing solutions primarily rely on modelcentric approaches based on entity-marked and entity-tagging representations. However, these methods require modifying the original text and model architecture, increasing complexity. Furthermore, they fail to provide adequate contextual and semantic information, leading to suboptimal performance, particularly in joint extraction settings. This research introduces TIRE, a data-centric framework that addresses these challenges through an innovative multi-sequence representation (MSR) for the RE. By incorporating key features such as Entity Mask and Entity Type, TIRE enhances contextual and semantic understanding, enabling precise classification of relationships between entities. Unlike complex model-centric approaches, TIRE achieves state-of-the-art performance with simplified architectures. Extensive evaluations on the curated DNRTI-AUG-STIX2-JE dataset demonstrate TIRE's superior performance in both pipeline and joint extraction settings, consistently achieving an F1 score of 99% in RE tasks while maintaining computational efficiency. TIRE's innovative design bridges the gap between NER and RE tasks for constructing high-quality cybersecurity knowledge graphs (CKGs) and shows adaptability to domains like finance, healthcare, and biomedical fields where structured information extraction is critical. This work underscores the potential of data-centric designs to advance relation extraction and support real-world applications.
Fetched live from OpenAlex and de-inverted. Abstracts are not stored in this database: the inverted indexes are 8.6 GB of the frame’s 9.3 GB of text, and the host has 13 GB free.
How this classification was reachedexpand
Full frame machine prediction
Teacher imitationNot calibrated prevalence, not ground truth. Human validation pending. The Gemma side is a direct model label for every work in the frame, read from the title-only record. The Codex side is a classifier learned from the 10,348 direct Codex labels and calibrated to design-weighted sample rates; fields without enough sample support carry no Codex call. Candidate is the union of the two sides; consensus is their intersection. These outputs are machine_predicted_unvalidated and are not human labels.
Distilled classifier scores by category (both heads)
| Category | Codex | Gemma |
|---|---|---|
| Metaresearch | 0.002 | 0.009 |
| Meta-epidemiology (narrow) | 0.002 | 0.001 |
| Meta-epidemiology (broad) | 0.001 | 0.003 |
| Bibliometrics | 0.007 | 0.006 |
| Science and technology studies | 0.001 | 0.001 |
| Scholarly communication | 0.003 | 0.009 |
| Open science | 0.003 | 0.004 |
| Research integrity | 0.002 | 0.003 |
| Insufficient payload (model declined to judge) | 0.004 | 0.005 |
Machine scores (provisional)
The two teacher heads of the student model, read on this work. A score orders the frame for review; it never asserts a category, and the validation status ships verbatim with every row.
Baseline scores from an immature model (maturity gate not passed, 7 training rounds). Scores rank; they never assert a category.
score_only:v0-immature-baseline · verbatim from the scoring run: score_only means the number may rank works, and no category label ships from itClassification
machine, unvalidatedMachine predicted; a candidate call from one source (direct Gemma or distilled Codex), not a consensus.
How this classification was reached, model by model and score by score, is at the end of the page under "How this classification was reached".