Dataset Overview

This project employs the CICIDS2017 network‑traffic dataset for machine‑learning‑based malicious traffic identification.

Libraries Used

• pandas – data manipulation, CSV I/O
• seaborn – data visualization
• matplotlib.pyplot – data visualization
• sklearn – machine‑learning algorithms
• numpy – numerical operations

Data Pre‑processing

1. Import Libraries – load pandas, seaborn, matplotlib, etc.
2. Read Data – use pandas.read_csv() to load CSV files for different dates.
3. Data Cleaning:
   • Handle missing values (e.g., drop rows or impute with mean/median).
   • Remove irrelevant columns.
   • Encode categorical features (e.g., label encoding).
   • Reduce memory usage (e.g., adjust dtypes).
   • Drop features with a single unique value.
4. Dimensionality Reduction (optional) – apply PCA or t‑SNE for visualization.
5. Feature Selection – analyze feature importance and select relevant features for model training.

Exploratory Data Analysis (EDA)

1. Distribution Analysis – use bar charts, histograms to explore feature distributions across traffic types (benign, DoS, etc.).
2. Correlation Analysis – heatmaps to identify relationships between features and target variable.
3. Class Imbalance – assess imbalance in traffic types; if present, apply oversampling techniques such as SMOTE.

Model Training

1. Data Split – use train_test_split to divide data into training and testing sets.
2. Model Selection – choose classification models (e.g., Random Forest). GPU‑accelerated options like cuml may be considered.
3. Training – train the model on the training set.
4. Evaluation – assess performance using accuracy, confusion matrix, classification report.

Results & Discussion

1. Present Results – report accuracy, confusion matrix, classification report.
2. Discuss Findings – analyze strengths, weaknesses, and impact of feature selection.
3. Future Work – outline potential improvements and further research directions.

Code Structure

• The code is organized by functionality: data preprocessing, feature engineering, model training, and evaluation, with comments explaining each section.