Verifier-Faithful In-Kernel Machine Learning for Network Intrusion Detection: generated results data

Generated results data supporting the article Verifier-Faithful In-Kernel Machine Learning for Network Intrusion Detection. Contains the canonical aggregator output (per_config_summary.csv and the per-corpus summaries), the raw per-trial outputs underlying every figure and table, the content-hashed calibration files, and the recorded execution environment. Seven public network-intrusion benchmarks are covered: NSL-KDD, UNSW-NB15, CIC-IDS2017, CSE-CIC-IDS2018, 5G-NIDD, ToN-IoT, and NF-UNSW-NB15-v2. Version 2.0.0 supersedes 1.0.0: all experiments were re-run at 30 repetitions per configuration on Linux 7.0.0-28, and the grid now emits per-class detection metrics (recall, false-positive rate, Matthews correlation, PR-AUC, and recall at a 1% false-positive budget) alongside accuracy and in-kernel faithfulness. Each results file declares STATUS: validated and the SHA-256 of the calibration under which it was produced. Confidence intervals are 5000-resample percentile bootstrap; pairwise comparisons carry Holm step-down correction. Source code is not part of this deposit. Raw third-party corpora are not redistributed; each is obtainable from the original host cited in the article's Data availability statement.