RNN-IDS-MODEL

LSTM IDS Models for CICIDS2017 Dataset

This repository provides two trained PyTorch LSTM models for network intrusion detection, trained on the CICIDS2017 dataset. These models are designed for use in research, benchmarking, or as a starting point for further development in network security and anomaly detection tasks.

Models Included

• lambda_with_valid.pth: LSTM model trained for binary classification (benign vs. attack) using cross-validation.
• mapping_with_valid.pth: LSTM model trained for multi-class attack categorization using cross-validation.

Model Details

• Architecture: LSTM-based Recurrent Neural Network
• Input Features: 80 per sample (preprocessed from CICIDS2017)
• Training: 5-fold cross-validation, early stopping, Adam optimizer
• Framework: PyTorch

Usage

1. Download the .pth files from this repository.
2. Load the model in your PyTorch code:

   import torch
   from your_model_definition import IdsRnn  # Use the same architecture as in training
   
   model = IdsRnn(hidden_size=512, output_size=2)  # or output_size=7 for multi-class
   model.load_state_dict(torch.load('lambda_with_valid.pth'))
   model.eval()
   

3. Prepare your input data with the same preprocessing as used during training.
4. Run inference as needed.

Notes

• These models require the same feature extraction and preprocessing pipeline as described in the original training code.
• For best results, refer to the full training pipeline and preprocessing steps.

License

MIT License

---

If you use these models in your research or project, please cite or reference this repository.