Data extraction dataset for a systematic literature review on prompt-based attacks and defenses in Large Language Models

This dataset supports a systematic literature review (SLR) on prompt-based attacks and defenses in Large Language Models (LLMs), following Kitchenham's guidelines and the PRISMA reporting framework. The dataset covers 89 primary studies published between 2023 and February 2026, retrieved from ArXiv, IEEE Xplore, and Scopus using the search string: ("large language model*" OR "LLMs") AND ("prompt injection" OR "jailbreak attack*" OR "prompt leak*" OR "malicious prompt*") AND (vulnerability OR mitigation). Files included: - data_extraction.csv: primary extraction spreadsheet with 89 studies and 31 fields covering attack techniques, defense strategies, threat models, evaluation metrics, and reproducibility indicators. - review_organization.csv: reviewer assignment sheet tracking Quality Assessment and Data Extraction completion per study and per reviewer. - data_extration_explore.ipynb: Jupyter notebook with exploratory data analysis of the extraction data. - README.md: full description of all fields and methodology.