lhmzll / deep-packet

Pytorch implementation of deep packet: a novel approach for encrypted traffic classification using deep learning

Deep Packet

Details in blog post: https://blog.munhou.com/2020/04/05/Pytorch-Implementation-of-Deep-Packet-A-Novel-Approach-For-Encrypted-Tra%EF%AC%83c-Classi%EF%AC%81cation-Using-Deep-Learning/

EDIT: 2022-01-18

• Update dataset and model

EDIT: 2022-01-17

• Update code and model
• Drop petastorm, use huggingface's datasets instead for data loader

How to Use

• Clone the project
• Create environment via conda
  • For Mac

    conda env create -f env_mac.yaml

  • For Linux (CPU only)

    conda env create -f env_linux_cpu.yaml

  • For Linux (CUDA 10.2)

    conda env create -f env_linux_cuda102.yaml

  • For Linux (CUDA 11.3)

    conda env create -f env_linux_cuda113.yaml

• Download the train and test set I created at here, or download the full dataset if you want to process the data from scratch.

Data Pre-processing

python preprocessing.py -s /path/to/CompletePcap/ -t processed_data

Create Train and Test

python create_train_test_set.py -s processed_data -t train_test_data

Train Model

Application Classification

python train_cnn.py -d train_test_data/application_classification/train.parquet -m model/application_classification.cnn.model -t app

Traffic Classification

python train_cnn.py -d train_test_data/traffic_classification/train.parquet -m model/traffic_classification.cnn.model -t traffic

Evaluation Result

Application Classification

Traffic Classification

Model Files

Download the pre-trained models here.

Elapsed Time

Preprocessing

Code ran on AWS c5.2xlarge

1:48:59 elapsed

Train and Test Creation

Code ran on AWS c5.2xlarge

1:26:27 elapsed

Traffic Classification Model Training

Code ran on AWS g5.xlarge

22:49 elapsed

Application Classification Model Training

Code ran on AWS g5.xlarge

7:41 elapsed