Pytorch implementation of GEE: A Gradient-based Explainable Variational Autoencoder for Network Anomaly Detection
Last updated Dec 9, 2025
29
Stars
12
Forks
1
Issues
0
Stars/day
Attention Score
18
Topics
Language breakdown
No language data available.
▸ Files
click to expand
README
GEE: A Gradient-based Explainable Variational Autoencoder for Network Anomaly Detection
Details in blog post: https://blog.munhou.com/2020/07/12/Pytorch-Implementation-of-GEE-A-Gradient-based-Explainable-Variational-Autoencoder-for-Network-Anomaly-Detection/
How to Use
Install Dependencies
Create a new conda environmentconda create -n gee python=3.7.7
conda activate gee
conda install pyspark=3.0.0 click=7.1.2 jupyterlab=2.1.5 seaborn=0.10.1
conda install pytorch=1.5.1 torchvision=0.6.1 cudatoolkit=10.1 -c pytorch
conda install pytorch-lightning=0.8.4 shap=0.35.0 -c conda-forge
pip install petastorm==0.9.2Feature Extraction
Download the processed data here or perform all the following steps.- Download raw data marchweek3csv.tar.gz and julyweek5csv.tar.gz.
- Decompress files.
tar -xvf marchweek3csv.tar.gz
tar -xvf julyweek5csv.tar.gz- Separate files by date.
grep '^2016-03-18' march.week3.csv.uniqblacklistremoved >> 20160318.csv
grep '^2016-03-19' march.week3.csv.uniqblacklistremoved >> 20160319.csv
grep '^2016-03-20' march.week3.csv.uniqblacklistremoved >> 20160320.csv
grep '^2016-07-30' july.week5.csv.uniqblacklistremoved >> 20160730.csv
grep '^2016-07-31' july.week5.csv.uniqblacklistremoved >> 20160731.csv- Put
20160319.csvand20160730.csvtodata/trainfolder,20160318.csv,20160320.csv, and20160731.csvtodata/testfolder. - Perform feature extraction.
python featureextraction.py --train data/train --test data/test --targettrain feature/train.feature.parquet --target_test feature/test.feature.parquetNormalise and Prepare Input Data for Model
Download the processed data here or perform all the following steps.python buildmodelinput.py --train feature/train.feature.parquet --test feature/test.feature.parquet --targettrain modelinput/train.modelinput.parquet --targettest modelinput/test.modelinput.parquetTrain Model
Download pre-trained model here or perform all the following steps.python trainvae.py --datapath modelinput/train.modelinput.parquet --model_path model/vae.model --gpu TrueEvaluation
ROC
Reconstruction Error Distribution
Gradient
