STACKWISE Empirical Evidence Benchmark: A Harmonised Multi-Source Dataset for Layer-Aware IoT Communication Stack Analysis

STACKWISE Empirical Evidence Benchmark v1.0.0 is a harmonised multi-source benchmark for reproducible analysis of IoT communication stacks under heterogeneous empirical evidence, measurement boundaries, uncertainty, and hard feasibility constraints. The benchmark is derived from four independently published real-world IoT measurement datasets covering NB-IoT/LTE-M energy consumption, WSN power consumption, LoRaWAN LR-FHSS energy measurements, and large-scale LoRaWAN gateway observations. The original raw datasets are not redistributed in this release. Instead, STACKWISE provides derived analysis-ready tables, canonical evidence records, provenance and measurement-boundary metadata, uncertainty and dependence descriptions, benchmark scenarios, candidate end-to-end communication stacks, hard-feasibility assessments, and clearly separated synthetic/model-derived sensitivity artifacts. Version 1.0.0 contains 398 canonical evidence records from four empirical sources, covering 14 metric families, seven benchmark scenarios, nine candidate communication stacks, and a complete 7 × 9 hard-feasibility matrix. The benchmark preserves the statistical unit, independence unit, derivation lineage, applicability domain, measurement boundary, and uncertainty semantics of the underlying evidence. The release is organised into six logical layers: L0 — Analysis-ready source derivatives: source-specific derived tables obtained from the published measurements.L1 — Canonical evidence records: harmonised cross-source evidence representation.L2 — Uncertainty: uncertainty, dependence, and robustness metadata.L3 — Benchmark definitions: scenarios, stack candidates, and component definitions.L4 — Feasibility and evidence support: hard-feasibility and decision-readiness information.L5 — Synthetic sensitivity: explicitly identified model-derived and synthetic sensitivity artifacts, kept separate from empirical measurements. The archive also contains schemas, source dataset cards, provenance and attribution information, release manifests, checksums, and machine-readable metadata intended to support independent reuse and verification. The benchmark does not claim to constitute newly collected primary measurements. It is a new harmonised derived research resource constructed from independently published empirical IoT datasets. It also does not provide a publication-grade global stochastic ranking of all candidate stacks or matched whole-device energy-per-report measurements across all cellular candidates. License: CC BY 4.0 for STACKWISE-authored benchmark material. Upstream datasets retain their respective original licenses and attribution requirements. Version: 1.0.0.