Machine Learning Cybersecurity Intrusion Detection

This repository contains code and data for detecting cyberattacks using machine learning techniques. The focus is on building and evaluating a Random Forest classifier to identify intrusion attempts based on network traffic data generated by the Cyber Range Lab at UNSW Canberra, Australia.

Model Performance:

The Random Forest classifier model to detect cyberattacks achieved the following performance metrics on the test set:

• Accuracy: 95.24% - High overall correctness.
• Recall: 93.07% - Effective in identifying actual attacks, minimizing false negatives.
• Precision: 98.99% - High precision, minimizing false positives.

The model's high precision and recall indicate strong performance in detecting cyberattacks, making it effective and reliable for real-world application.

Data Analysis

• Conducted correlation analysis with heatmaps to identify network variables associated with cyberattacks.
• Identified network variables positively and negatively correlated with cyberattacks and highlighted strong correlations among variables.
• Evaluated model performance and visualized feature importances.