An Enhanced Detection of DDoS Using Random Forest with Recursive Feature Elimination and Tree of Parzen Estimators

Model Architecture

1. Original DDoS Dataset Analysis

Workflow:

• Data Collection
• Exploratory Data Analysis: Shape, Distribution of Class & Label, Outlier Detection
• Data Pre-Processing: Handling Nulls, Duplicates, Low-Variance Columns; Data Splitting; Data Scaling; Feature Selection using Recursive Feature Elimination (RFE)
• Model Training using Random Forest
• Hyperparameter Optimization using Tree of Parzen Estimators (TPE)
• Model Evaluation: Achieved 99.5% accuracy with unbalanced dataset

2. Data Augmentation

Steps:

• Environment & Hardware Check
• Data Loading
• Metadata Creation & Saving
• Synthetic Data Generation using Conditional Tabular Generative Adversarial Network (CTGAN)
• Model Saving into .csv format

3. Synthesized DDoS Dataset Analysis

Workflow:

• Data Collection
• Exploratory Data Analysis: Shape, Distribution of Class & Label, Outlier Detection
• Data Pre-Processing: Handling Nulls, Duplicates, Low-Variance Columns; Data Splitting; Data Scaling; Feature Selection using Recursive Feature Elimination (RFE)
• Model Training using Random Forest
• Hyperparameter Optimization using Tree of Parzen Estimators (TPE)
• Model Evaluation: Achieved 96.4% accuracy with unbalanced dataset

In detail Document scan the following QR Code