A security data scientist is building a network intrusion detection model. She splits her labeled dataset into three subsets before training. What is the correct purpose of each split?

• A.All three splits are used simultaneously during the gradient descent optimization process
• B.Train and Validation are used identically; Test is used for model training; datasets serialized as Parquet physically prevent any row from being read during both training and evaluation, making train-test leakage a storage-layer impossibility rather than a methodology concern
• C.Train: initial testing; Validation: model building; Test: final deployment dataset; an air-gapped deployment makes a model immune to adversarial inputs, since crafted examples require live network access to the weights
• D.Train: the model learns patterns from this set; Validation: used during training to tune hyperparameters and detect overfitting; Test: held-out final evaluation on unseen data to estimate real-world performance