munhyunsu/tff-app

Federated learning based app

Federated Learning Research Meterials

[2022 FL-AAAI] FLHub: Federated Learning Hub

[2021 KSC] gRPC-based Federated Learning Model Service

FLHub: Federated Learning Hub

• Prototyping for FLHub

Service components

• FLHub server
• FLHub CLI client

[2020 MDPI-Electronics] Traffic Classification using Federated Learning Simulation

Experiment list

1. Centralized vs. Federated
2. The number of clients
3. Data type: Balanced / NonIID

TODO

• Protobuf generator

  python3 01-create_protobuf.py

• Dynamic classification model

  • Model ckpt saver / restorer

Requirements

• PcapPlusPlus

Packet to image (pkt2vec)

python3 00-Preprocessing.py

• Image with 375 x 4 x 1 from ip layer

Packet Classification Baseline

02-TCBaselineCNN.ipynb

• Models

_________________________________________________________________
Layer (type)                 Output Shape              Param #
=================================================================
conv2d_1 (Conv2D)            (None, 39, 37, 32)        320
_________________________________________________________________
flatten_1 (Flatten)          (None, 46176)             0
_________________________________________________________________
dense (Dense)                (None, 15)                692655
=================================================================
Total params: 692,975
Trainable params: 692,975
Non-trainable params: 0

Federated Server

Requirements

pip3 install --upgrade -r requirements.txt

• Install scrcpy

Usage

1. Preprocessing pcap files for creating feature vectors

python3 01-preprocess.py -i INPUTDIR -o OUTPUTDIR

2. Make dataset for federated learning

- 

python3 02-...

Edge Client

Requirements

pip3 install -r requirements.txt

Usage

1. ...

References

Federated Optimization: Distributed Optimization Beyond the Datacenter Communication-Efficient Learning of Deep Networks from Decentralized Data

Coded by

• LuHa (munhyunsu@gmail.com)