A Risk-Aware Access Control Model Based on Multi-Attribute Behavior Modeling and Fuzzy Inference

Dataset Description for Zenodo This dataset contains the data files used for the experimental evaluation of the proposed risk-aware access control model based on multi-attribute behavior modeling and fuzzy inference. The data/ directory includes controlled experimental data, processed datasets, semi-real cybersecurity data, UNSW-NB15-based validation data, and LANL authentication-log-based validation data. These datasets are used to support reproducible experiments, including baseline comparison, ablation analysis, sensitivity analysis, semi-real validation, and real authentication-log validation. Processed Experimental Data The data/processed/ directory contains processed datasets that were generated from the controlled data generator, the semi-real cybersecurity dataset, the UNSW-NB15 dataset, and the LANL authentication-log sample. These files are provided to support direct reproduction of the experimental results without requiring users to repeat every raw-data mapping or preprocessing step. The processed datasets include: Controlled access-control experiment data generated by the labeled data generator. Train/test splits for the main controlled-data experiments. Processed semi-real cybersecurity validation data. Processed UNSW-NB15 validation data mapped into access-control-related behavioral risk features. Processed LANL authentication-log validation data mapped from auth_sample_2gb.txt and redteam.txt. In these processed files, the original raw or semi-raw records have been transformed into a unified risk-aware access-control format. The main fields include user/request identifiers, behavioral risk indicators, normalized feature values, risk labels, and train/test partitions used by the experiments. The processed files are used by the experiment scripts to reproduce baseline comparison, ablation analysis, sensitivity analysis, external validation, and diagnostic experiments. When the original large raw datasets are unavailable, the processed files can still be used to reproduce the main experimental evaluations reported in the manuscript. LANL Authentication Log Data The LANL authentication-log sample file data/LANL/auth_sample_2gb.txt was derived from the LANL Cyber Security Dataset, specifically the authentication log (auth.txt) available from: https://csr.lanl.gov/data/cyber1/ Because the original LANL authentication log is very large, this repository does not use the full raw file directly. Instead, auth_sample_2gb.txt was generated by randomly sampling approximately 2 GB of valid authentication records from the original LANL auth.txt file. During sampling, only complete and valid records were retained. Records containing missing values, question marks (?), blank fields, or incomplete field structures were excluded. The resulting sampled file preserves complete LANL authentication-event records and is used as the real authentication-log dataset for supplementary validation experiments. The LANL red-team annotation file data/LANL/redteam.txt is used together with auth_sample_2gb.txt to identify known malicious or compromised authentication-related events and to construct risk labels for access-control evaluation. UNSW-NB15 Data The UNSW-NB15 validation data included in data/UNSW_NB15_CSV/ was derived from the UNSW-NB15 dataset, which is publicly available from the following official source: https://research.unsw.edu.au/projects/unsw-nb15-dataset In this dataset package, the UNSW-NB15 training and testing files are used to construct a semi-real network-security validation setting for risk-aware access control evaluation. The original UNSW-NB15 traffic records were mapped into access-control-related behavioral risk features and risk labels, enabling comparison between the proposed method and baseline methods under a publicly available cybersecurity dataset. Usage This dataset package is intended for academic research and reproducibility purposes. Users should cite or acknowledge the original LANL Cyber Security Dataset and UNSW-NB15 Dataset, and comply with the usage terms of the original data sources.