Smart Anomaly Detection and Threat Analysis System

I’m putting together an academic-grade project that demonstrates how supervised learning can spot common web attacks—specifically SQL Injection, Cross-Site Scripting (XSS), and Distributed Denial-of-Service (DDoS). I will supply a custom dataset; your task is to build, train, and test a Python solution that flags these threats in near-real time.

Here’s the flow I have in mind:

• Clean and prepare the dataset, outlining the feature-engineering decisions so the process can be repeated or extended later.
• Train a supervised model (classical algorithms such as Random Forest, SVM, or a lightweight deep-learning variant if it boosts accuracy). Explain why the chosen approach suits multi-class attack detection.
• Evaluate performance with accuracy, precision, recall, and confusion matrices, then include short commentary on any class-imbalance handling.
• Package the inference pipeline so it runs from the command line and a simple Jupyter notebook, making it easy for examiners to execute and see live predictions.
• Provide well-commented source code, a requirements.txt, and step-by-step instructions that go from environment setup to final evaluation.

Deliverables:
1. Full Python source and trained model files
2. The processed version of the dataset plus scripts that generate it from raw data
3. A concise PDF explaining logic, workflow, and how to rerun everything (screenshots welcome)
4. Optional extras: a short report (~10 pages) and a slide deck to support a final-year viva—include these if you’re comfortable producing them, otherwise let me know so we can adjust milestones accordingly.

Keep the solution native to widely used libraries (pandas, scikit-learn, TensorFlow/Keras or PyTorch) so reviewers can reproduce results without exotic dependencies. Accuracy matters, but clarity and reproducibility are paramount—I want to be able to hand this over, have someone install requirements, press “Run,” and immediately understand each stage of the pipeline.