Phishing Detection

This project utilizes the Gated Recurrent Unit, Long Short-Term Memory, Multi-Layer Perceptron, and Transformer Encoder models to detect phishing emails.

Installation

Navigate to the project directory and install the package in editable mode

cd /PhishDetect
pip install -e .

Alternatively, if you prefer to work with local imports:

pip install yacs einops torch numpy pandas scikit-learn transformers

Optional Packages

Although not necessary, Weights and Biases (wandb) is a good tool for tracking training progress:

pip install wandb
wandb login

Usage

After installing the codebase as a package, you can now use PhishDetect anywhere:

import PhishDetect.utils.arguments as args
import PhishDetect.models
import PhishDetect.datasets

Running the Models

Training:

python main.py --cfg settings/config_files/MODEL_TYPE.yaml --opts EXECUTION.MODE train

"train.sh" also provides the functionality of training all four model types at once.

Demo:

python main.py --cfg settings/config_files/MODEL_TYPE.yaml --opts EXECUTION.MODE demo