PhishNet: phishing website classification
PhishNet is an experimental PyTorch classifier for phishing websites using six engineered numeric website features. This repository shares the saved model weights from Lakshitha Perera's GitHub project.
What is included
best_model.pt: thestate_dictsaved when validation loss improved in the original training script.- Source code, notebooks, and dataset on GitHub.
The source network has fully connected layers of 6 → 64 → 32 → 2 with ReLU activations. The two outputs are classification logits. See the training script for the exact feature preparation and training workflow.
Important limitations
This is a project artifact, not a ready-to-use URL scanner. The checkpoint does not include the fitted StandardScaler or a URL feature extraction pipeline. The source script fits its scaler before the train/validation/test split, so reported evaluation results may be optimistic. Refit preprocessing on training data only and validate on held-out data before using this model in any security workflow. Do not use this experimental model as the sole basis for a security decision.
Source
GitHub: https://github.com/lakshitha1629/PhishNet_Deep_Learning_for_Phishing_Detection