ClieND

Repo for "ClieND: Client-Side Neuron-Level Detection against Poisoning Attacks on Cross-Silo Federated Learning" in Asia CCS'26.

Contents

• config.json — Experiment configuration (datasets, model, FL/attack settings).
• utils_models.py, data_utils.py, attacks.py, training.py, utils_file.py — Model, data loading, attack, training, and I/O utilities.
• main.py (or your entry script) — Federated training with optional attacks; logs and results are saved under results/.

Quick Start

# 1) create env and install deps
#    Python ≥3.10, PyTorch ≥2.1 recommended
#    pip install torch torchvision pandas

# 2) Edit config.json to choose dataset/model and attack settings

# 3) Run
python main.py