Reproducibility Evidence for Limits of Robust Aggregation: A Cross-Attack and Cross-Dataset Evaluation
Description
Audited derived evidence supporting the manuscript ‘Limits of Robust Aggregation: A Cross-Attack and Cross-Dataset Evaluation of Poisoning Defences for Federated IoT Intrusion Detection.’ The deposit contains completion audits, run-level summaries, primary-test rows, controlled GPU timing rows, client-risk source rows, reproducibility scripts, selected tables, figures, and checksums. The underlying N-BaIoT and TON_IoT benchmark datasets are publicly available and are not redistributed in this deposit.
Files
Steps to reproduce
Extract Research_Data_Deposit_Package.zip. Review the included README and integrity/checksum files first. The package contains the audited broad- and focused-archive completion records, derived run-level summaries, primary-test rows, controlled GPU timing evidence, client-risk source data, selected tables and figures, and reproducibility scripts. Run the supplied rebuild/reproduction script in the environment described in the package to regenerate the corrected statistical summaries, selected tables, and figures from the included CSV evidence. The package supports verification of the reported derived results but does not claim to rerun the complete federated-training experiments from raw datasets. The original N-BaIoT and TON_IoT benchmark datasets are publicly available from their respective sources cited in the manuscript.