chinmayee-04/NeuroGuard-Adaptive-Cognitive-Threat-Mitigation-System

# NeuroGuard: Adaptive Cognitive Threat Mitigation System     ## Overview NeuroGuard is a real-time network intrusion detection system trained on the KDD Cup dataset. It applies multiple machine learning and deep learning models to classify network traffic as normal or attack, with real-time traffic simulation and automated threat response. ## Dataset - **KDDTrain+.txt** - NSL-KDD dataset with 43 network traffic features - Labels classified as: `normal` or `attack` - Features include: duration, protocol_type, service, flag, src_bytes, dst_bytes, and 38 additional network-level attributes ## Models Implemented | Model | Type | |-------|------| | Logistic Regression | Classification | | K-Nearest Neighbors (k=20) | Classification | | Gaussian Naive Bayes | Classification | | Linear SVC | Classification | | Decision Tree | Classification | | Random Forest | Classification (Best Model) | | XGBoost | Regression | | Deep Neural Network (TensorFlow) | Binary Classification | | PCA + Random Forest | Dimensionality Reduction + Classification | ## Workflow 1. Data Loading - Load KDDTrain+.txt from Google Drive 2. Preprocessing - Assign column names, encode labels, scale features using RobustScaler 3. EDA - Visualize protocol types and attack distribution using pie charts 4. Feature Engineering - Apply PCA (20 components) for dimensionality reduction 5. Model Training - Train and evaluate all models, compare accuracy, precision, recall and F1 score 6. Deep Learning - Build and train a Dense Neural Network using TensorFlow/Keras 7. Best Model Saved - Random Forest saved as `best_random_forest_model.pkl` 8. Real-Time Simulation - Simulate live traffic using PCA-transformed test data and predict threats row by row ## Tech Stack | Category | Tools | |----------|-------| | Language | Python 3.9+ | | Deep Learning | TensorFlow, Keras | | Machine Learning | Scikit-learn, XGBoost | | Data Processing | Pandas, NumPy | | Visualization | Matplotlib, Seaborn | | Dimensionality Reduction | PCA (20 components) | | Model Saving | Joblib | | Platform | Google Colab | | Dataset | NSL-KDD (KDDTrain+.txt) | ## Repository Structure NeuroGuard/ ├── notebooks/ │ └── MCSA__ipynb.ipynb ├── dataset/ │ └── KDDTrain+.txt ├── models/ │ └── best_random_forest_model.pkl ├── report/ │ └── final_report_42614152.pdf ├── research/ │ └── Neuroguard_research_paper.pdf ├── README.md └── requirements.txt