Trustworthy Federated Learning for Cybersecurity: A Three-Level Assurance Taxonomy, Design-Space Audit, and Source-Linked Critical Review.

Abstract

This repository contains the supporting evidence and reproducibility materials for the manuscript “Trustworthy Federated Learning for Cybersecurity: A Three-Level Assurance Taxonomy, Design-Space Audit, and Source-Linked Critical Review.” The package includes: A study-feature matrix covering 20 studies and 18 features. A source-linked audit ledger containing 360 study-feature judgments, with source references, evidence locators, and coding notes. Feature definitions, coding guidance, and mappings between the original and current study identifiers. Supporting Information in PDF and LaTeX formats. Python scripts for generating figures, preparing data partitions, and checking the experimental implementation. Documentation, validation results, and file checksums. These materials allow reviewers and researchers to inspect the recorded evidence, understand the coding decisions, and reproduce the documented data-derived outputs. The package contains literature-review data and supporting code; it does not include third-party raw cybersecurity datasets. Synthetic examples are provided for software checks and do not represent real-data experimental findings. Internal consistency checks do not constitute independent re-coding or fresh verification of every cited source. The README explains the package’s provenance, usage, scope, and remaining limitations.