LSTM DDoS Detection for SDN-VANET

I hold a sizeable SDN-VANET traffic dataset (currently in a proprietary/“other” format) and need it transformed into a working LSTM-based DDoS detector. The core of the job is to take an existing Long Short-Term Memory architecture, adapt it to the dataset, and demonstrate reliable attack detection.

The flow I have in mind is straightforward:

• Parse the raw dataset and convert or load it into a format that common deep-learning libraries (TensorFlow, Keras, or PyTorch) can read.
• Apply full pre-processing—data cleaning, feature extraction, and normalization—so the sequences feed cleanly into the network.
• Implement the chosen LSTM model, train, validate, and test it, then report key metrics such as accuracy, precision/recall, F1, and any relevant confusion-matrix insights for DDoS versus benign traffic.
• Package all code, notebooks, and a concise read-me so I can reproduce results locally or on a GPU instance.

Feel free to suggest minor hyper-parameter tweaks or regularisation techniques if they improve performance, but the backbone must remain LSTM. If you have experience with SDN, VANET, or network-intrusion datasets, that will be a plus.

I’m ready to start as soon as you confirm you can handle the data conversion step and provide an estimated timeline for each stage.