Calibrated, Lead-Time-Aware Exploitability Prediction: Dataset Snapshot and Reproducible Pipeline

Complete reproducibility artifact for the paper "Calibrated, Lead-Time-Aware Exploitability Prediction: Probabilities and Warning Time for Vulnerability Triage" (Ali Abdullah S. AlQahtani, Prince Sultan University). This archive contains the full code pipeline, a frozen point-in-time data snapshot, and the figures needed to reproduce every result in the paper. The dataset is derived from four public sources — the CISA Known Exploited Vulnerabilities catalog (labels), historical EPSS scores, NVD vulnerability metadata, and time-gated Metasploit module metadata — joined on (CVE, observation date) under a strict temporal split that forbids look-ahead. All headline results, including baseline model sweep, case-control ratio sensitivity, rolling-origin cross-validation, base-rate-corrected ranking, calibration, conformal coverage, and the temporal-leakage analysis, regenerate offline by running run_all.sh. No synthetic data or model-generated labels are used. Code is released under the MIT License; the redistributed data remains under the terms of its upstream sources.