Analyzing Deep Learning Model Performance for Intrusion Detection on CIC-IDS2017 Dataset

This repository contains the research paper titled "Analyzing Deep Learning Model Performance for Intrusion Detection on CIC-IDS2017 Dataset" by Md. Sadman Sakib and Noshin Tabassum, published as part of our research.

📄 Abstract

Intrusion Detection Systems (IDS) are vital for detecting and mitigating real-time cyber threats. This study leverages the CIC-IDS2017 dataset to evaluate the performance of deep learning models—including MLP, 1D CNN, LSTM, BiLSTM, GRU, DBN, and hybrid architectures—for both binary and multiclass intrusion classification tasks. Our findings show that deep learning significantly improves intrusion detection accuracy and reduces false positives compared to conventional machine learning models.

🧠 Models Used

• Multi-Layer Perceptron (MLP)
• 1D Convolutional Neural Network (CNN)
• Long Short-Term Memory (LSTM)
• Gated Recurrent Unit (GRU)
• Bidirectional LSTM (BiLSTM)
• Deep Belief Network (DBN)
• CNN + LSTM
• CNN + BiLSTM

🧪 Dataset

• CIC-IDS2017 Dataset
  Provided by the Canadian Institute for Cybersecurity.

🛠 Tools & Environment

• Google Colab
• Python 3.x
• TensorFlow / Keras / PyTorch
• Scikit-learn, NumPy, Pandas, Matplotlib, Seaborn

📊 Evaluation Metrics

• Accuracy
• Precision
• Recall
• F1 Score
• Training Time

📘 Notebook

The full code—including preprocessing, skewness correction, deep learning model training, and evaluation—is available in:

📄 notebook/deep_learning_ids_cicids2017.ipynb

📎 Paper

You can find the full PDF of the paper in this repository: Paper.pdf

📬 Contact

Md. Sadman Sakib
Department of ECE, KUET
📧 [email protected]
🔗 LinkedIn | GitHub