The source code and models for our paper AEGIS-Net.
All code was developed and tested on a single machine equiped with a Tesla V100 32GB GPU. The environment is as bellow:
- Ubuntu 20.04.6 LTS
- Python 3.10.13
- Pytorch 2.1.1
- faiss-gpu==1.7.2
- numpy==1.26.1
- pandas==2.1.3
The code is currently tested only on GPU.
-
Data preparation
Preprocessing of the dataset is done at
Data_Proprecessing.Created a folder
AEGIS-Net/datasetand downloadMalicious_TLS/CICIDS2017/CICIDS2017dataset into this folder.You can download these datasets for pan.baidu.com.
AEGIS-Net dataset Link: https://pan.baidu.com/s/14dWEJquGwDoPuyt5zej4jw The extraction code: hrv7 CIC_IDS_2017_Formated_DATA.csv is the CICIDS2017 dataset. processed_friday_dataset.csv is the CICIDS2018 friday dataset(2018.3.2 and 2018.2.16). cleaned_data is the CICDarknet2020 dataset.
-
Source code
# Execute the following three files in sequence
1.python training.py
2.python extract_features.py
3.python Unknown_detection.py# Run at different noise rate
sh training.sh