SDN VANET Simulation & ML Detection
Budget: ₹1,500 – ₹12,500 INR
I need a working, reproducible setup that lets me study DDoS threats in an SDN-enabled Vehicular Ad-Hoc Network. Here is the scope I have in mind:
• Environment – Build a Mininet-WiFi topology that links to SUMO for realistic vehicle mobility. Include the controller scripts I can drop straight into the SDN controller of choice (Ryu or something similar) so the network boots with the right flows and can be re-run on demand.
• Traffic generation – Script both benign and DDoS traffic. I did not lock myself into a single pattern, so feel free to suggest CBR, VBR, Poisson or a mix if it helps produce a balanced dataset.
• Dataset output – Capture cross-layer features (MAC, network, transport and flow-level stats) and export them to CSV/Parquet so they are ready for analysis.
• Detection models – Using scikit-learn, train and evaluate at least two machine-learning classifiers that can tell DDoS from normal traffic. Provide the training notebooks / .py files plus a concise results summary (accuracy, precision, recall and confusion matrix).
• Handover – Deliver all source code, a requirements.txt, clear setup steps, and a short write-up describing how to rerun the simulation, regenerate the datasets and reproduce the ML results.
If something in the outline can be improved without blowing up complexity, I’m open to your suggestions.
• Environment – Build a Mininet-WiFi topology that links to SUMO for realistic vehicle mobility. Include the controller scripts I can drop straight into the SDN controller of choice (Ryu or something similar) so the network boots with the right flows and can be re-run on demand.
• Traffic generation – Script both benign and DDoS traffic. I did not lock myself into a single pattern, so feel free to suggest CBR, VBR, Poisson or a mix if it helps produce a balanced dataset.
• Dataset output – Capture cross-layer features (MAC, network, transport and flow-level stats) and export them to CSV/Parquet so they are ready for analysis.
• Detection models – Using scikit-learn, train and evaluate at least two machine-learning classifiers that can tell DDoS from normal traffic. Provide the training notebooks / .py files plus a concise results summary (accuracy, precision, recall and confusion matrix).
• Handover – Deliver all source code, a requirements.txt, clear setup steps, and a short write-up describing how to rerun the simulation, regenerate the datasets and reproduce the ML results.
If something in the outline can be improved without blowing up complexity, I’m open to your suggestions.
Related categories:
Python
Linux
Machine Learning (ML)
Network Administration
Data Analysis
Network Security
Simulation
Mininet