Dataset Overview

Dataset Information

Dataset Name

• CICIDS2018 Dataset

Dataset Description

• Description: This dataset contains labeled network traffic data covering multiple attack types (e.g., DoS, brute‑force, SQL injection, botnet) and normal traffic.
• Link: The dataset can be downloaded here.
• Size: Large dataset split into multiple CSV files, total size exceeds several hundred MB.

Dataset Usage

• Training Data: dataset/train_data.csv
• Test Data: dataset/test.csv
• Training Data Version: artifacts/train_data.csv

Dataset Processing

Data Ingestion

• Script: src/components/data_ingestion.py

Data Transformation

• Script: src/components/data_transformation.py

Model Training

• Script: src/components/model_trainer.py

Model Performance

Test Accuracy

• Test Accuracy: 89.75%
• Training Accuracy: 89.87%

F1 Score

• Test F1 Score: 88.27%
• Training F1 Score: 88.40%

Recall

• Test Recall: 89.75%
• Training Recall: 89.87%

Precision

• Test Precision: 89.08%
• Training Precision: 89.31%

Balanced Accuracy

• Balanced Accuracy: 86.55%

ROC AUC

• Test ROC AUC: 99.17%
• Training ROC AUC: 99.21%