Reproducible Synthetic Smart-Campus Federated Learning Simulation: Code and Results for PPFL-SC

This repository provides the synthetic experimental data, Python code, and reproducibility materials supporting the manuscript entitled "Engineering Privacy-Aware Federated Learning for Smart Campus Services." The experimental package includes a synthetic federated learning simulation with 5,000 training samples, 2,000 test samples, and 10 simulated clients. The experiments compare conventional FedAvg with sparse error-feedback updates and optional Gaussian perturbation under different data heterogeneity conditions. The repository contains source code, experimental configurations, random seeds, per-round results, summary statistics, and supporting documentation. All experimental data are synthetically generated. No real student records, institutional datasets, or human-participant information were collected or processed. The simulation provides preliminary computational evidence and does not establish real-world deployment performance or a formal differential privacy guarantee. OpenAI ChatGPT provided substantive assistance in developing the simulation code and documentation. The associated manuscript contains an AI-use disclosure.