Reproducibility Package for PG-FRIDS: Resource-Aware Lightweight Security Analytics across the Cloud-Edge Continuum
Description
This dataset provides the reproducibility package supporting the manuscript “Pareto-Guided Resource-Aware Selection of Lightweight Security Analytics for Resource-Constrained Edge Tiers of the Cloud-Edge Continuum.” It contains derived experimental results, dataset provenance and split-integrity records, training-only feature-selection rankings, repeated software-level resource-profiling outputs, Pareto and resource-priority mapping results, model-configuration records, environment and protocol metadata, supporting analysis scripts, sensitivity checks, and manuscript-synchronised figures. The experiments use the publicly available UNSW-NB15 and ToN_IoT network datasets. Raw copies of these public datasets are not redistributed in this repository. UNSW-NB15 current numerical authority is based on the canonical v36 rerun, while ToN_IoT results remain frozen from v35. The package is intended to support verification and reproducibility of the reported software-level resource-aware benchmarking results. It does not constitute heterogeneous-hardware deployment validation and does not report direct energy or carbon measurements.
Files
Steps to reproduce
Obtain the public UNSW-NB15 and ToN_IoT network datasets from their authoritative UNSW Canberra sources. Raw source datasets are not redistributed in this repository. Follow the dataset provenance, split-integrity, preprocessing, and environment records supplied in the reproducibility package. Feature selection is fitted on the training data only and the resulting feature indices are subsequently applied to the corresponding test data. Use the supplied model-configuration and environment records to reproduce the lightweight classifier baselines and reduced-feature configurations. Run the provided resource-profiling workflow using the documented warm-up, repetition, batch-prediction, model-size, memory, latency, and throughput protocol. Use the supplied feature-reduction and Pareto-analysis scripts to reproduce the feature-reduced results, Pareto-efficient candidate sets, and resource-priority mappings. Compare regenerated outputs with the supplied derived-result tables, profiling records, manifests, and checksums. The canonical UNSW-NB15 numerical authority corresponds to the v36 rerun; ToN_IoT results remain frozen from v35.
Institutions
- National University of MalaysiaSelangor, Bangi