The script:

• removes duplicates and empty lines,
• cleans texts (spaces, line breaks, tabs),
• normalizes labels, and it generates:
• data/combined.csv: cleaned complete dataset,
• data/train.csv: training data (80%),
• data/test.csv: test data (20%).

3. Model training
   root@kitploit:~

   python scripts/train_model.py
   

The model uses an ML pipeline:

TF-IDF Vectorizer (TfidfVectorizer): transforms text requests into numerical vectors, capturing the importance of words or suspicious symbols.

Logistic Regression (LogisticRegression): supervised classifier that learns to associate TF-IDF vectors with the classes sql_injection, xss or benign.

After training, the model is saved in the models folder; it can then be used to predict the class of a new request.

4. Automatic tests
   root@kitploit:~

   pytest -v -s tests/test_pipeline.py
   

The test verifies that:

• the model is loadable,
• it correctly predicts several simple examples,
• no pipeline error occurs.

5. Python dependencies
   root@kitploit:~

   pip3 install -r requirements.txt
   

Libraries

• pandas
• scikit-learn
• joblib
• pytest

6. Notes
   The scripts are designed to be run locally, in a Python 3.13+ environment with the listed dependencies.