FedIoT: Federated Learning for Internet of Things

This repository is the official implementation of Federated Learning for Internet of Things: A Federated Learning Framework for On-device Anomaly Data Detection. Read our paper here: https://arxiv.org/abs/2106.07976

1. Introduction

Due to the heterogeneity, diversity, and personalization of IoT networks, Federated Learning (FL) has a promising future in the IoT cybersecurity field. As a result, we present the FedIoT, an open research platform and benchmark to facilitate FL research in the IoT field. In particular, we propose an autoencoder based trainer to IoT traffic data for anomaly detection. In addition, with the application of federated learning approach for aggregating, we propose an efficient and practical model for the anomaly detection in various types of devices, while preserving the data privacy for each device. What is more, our platform supports three diverse computing paradigms: 1) on-device training for IoT edge devices, 2) distributed computing, and 3) single-machine simulation to meet algorithmic and system-level research requirements under different system deployment scenarios. We hope FedIoT could provide an efficient and reproducible means for developing the implementation of FL in the IoT field.

Check our slides here: https://docs.google.com/presentation/d/1aW0GlOhKOl35jMl1KBDjKafJcYjWB-T9fiUsbdBySd4/edit?usp=sharing

2. Installation

After git clone-ing this repository, please run the following command to install our dependencies.

conda create -n fediot python=3.7
conda activate fediot
pip install torch==1.6.0+cu101 torchvision==0.7.0+cu101 -f https://download.pytorch.org/whl/torch_stable.html
pip install -r requirements.txt 
git submodule add https://github.com/FedML-AI/FedML
cd FedML; git submodule init; git submodule update; cd ../;

For the FedML package installation, please check http://doc.fedml.ai/#/installation-distributed-computing

3. Data Preparation

Run sh download.sh under data/UCI-MLR to download dataset. We select the N_BaIoT data as our evaluation dataset. For the detailed of our data, please look at the data_readme.md in the data folder.

4. Code Structure of FedIoT

FedML: a soft repository link generated using git submodule add https://github.com/FedML-AI/FedML.
data: provide data downloading scripts and store the downloaded datasets. Note that in FedML/data, there also exists datasets for research, but these datasets are used for evaluating federated optimizers (e.g., FedAvg) and platforms. FedNLP supports more advanced datasets and models.
data_preprocessing: data loaders, partition methods and utility functions
model: IoT models. For example, VAE for outlier detection.
training: please define your own trainer.py by inheriting the base class in FedML/fedml-core/trainer/fedavg_trainer.py. Some tasks can share the same trainer.
experiments/distributed:

experiments is the entry point for training. It contains experiments in different platforms. We start from distributed.
Every experiment integrates FOUR building blocks FedML (federated optimizers), data_preprocessing, model, trainer.
To develop new experiments, please refer the code at experiments/distributed/main_uci_vae.py.

experiments/Raspberry Pi:

It is the code designed for the implementation on the Raspberry Pi 4b.
It contains two blocks, main_uci_rp.py should be implemented on the edge device and app.py should be implemented on the server.
For the more detailed running setup, please look at the Resberry_Pi Readme.md at the experiments/Resberry_Pi.

experiments/centralized:

please provide centralized training script in this directory.
This is used to get the reference model accuracy for FL.
You may need to accelerate your training through distributed training on multi-GPUs and multi-machines. Please refer the code at experiments/centralized/ae_cen_glb_test.py.

5. Results

Please read the experiment section in our paper. The main function under the distributed now sets only for model training, please use the fl_test to evaluate the performance.

Please note that you should change the aggregator w = 1/9 for a better result. It could be changed in FedML/fedml_api/distributed/fedavg/FedAVGAggregator.py line 77

6. Citation

Please cite our FedIoT and FedML paper if it helps your research. You can describe us in your paper like this: "We develop our experiments based on FedIoT [1] and FedML [2]".

@article{Zhang2021FederatedLF, title={Federated Learning for Internet of Things: A Federated Learning Framework for On-device Anomaly Data Detection}, author={Tuo Zhang and Chaoyang He and Tian-Shya Ma and Mark Ma and S. Avestimehr}, journal={ArXiv}, year={2021}, volume={abs/2106.07976} }

7. Contact

The corresponding author is:

Tuo Zhang tuozhang@usc.edu

Chaoyang He chaoyang.he@usc.edu http://chaoyanghe.com

Special Thanks to Tianhao Ma!

S16201512/FedIoT