Installation Guide

Installation Guide

Requirements

• Python 3.9 or above
• Jupyter Notebook or JupyterLab
• Minimum 4 GB RAM (8 GB recommended)
• Internet connection for pip installation

Step 1: Extract the Project

Extract the downloaded ZIP file to a folder on your system. The project folder should be named Voice phishing (vishing) and contain training.ipynb, app.py, the fake_audio_dataset folder, and the templates folder.

Step 2: Open Terminal or Command Prompt

Navigate to the project folder using your terminal or command prompt.

cd "Voice phishing (vishing)"

Step 3: Install Dependencies

Run the following command to install all required Python packages.

pip install -r requirements.txt

If you are on Apple Silicon (M1, M2, or M3), open requirements.txt, remove the tensorflow line, and install the macOS-compatible version instead.

pip install tensorflow-macos tensorflow-metal

If you have an NVIDIA GPU and want CUDA acceleration, use the following instead.

pip install tensorflow[and-cuda]

Step 4: Open the Training Notebook

Launch Jupyter Notebook from the project folder.

jupyter notebook training.ipynb

Step 5: Run All Cells

In Jupyter, go to the menu and click Kernel, then click Restart and Run All. Training takes between 5 and 20 minutes depending on your hardware. Wait until you see a message similar to the following in the output of the last cell.

Best model: CNN+LSTM  (F1=0.9333)
Model saved to best_model.h5
Config saved to config.pkl

Step 6: Verify Training Output

After training completes, confirm that the following files exist in the project root folder.

• best_model.h5
• config.pkl
• cnn_best.h5
• cnn_lstm_best.h5

Step 7: Launch the Web Application

Open a new terminal window, navigate to the project folder, and run the Flask application.

python app.py

Step 8: Open in Browser

Open your browser and go to the following address.

http://127.0.0.1:5001

The home page should load. Go to the Detect page, upload an audio file or record using the microphone, and click the detect button to get a prediction.

Supported Audio Formats

• .wav
• .mp3
• .flac
• .ogg
• .m4a
• .webm (browser recording)

Common Issues

best_model.h5 not found

This means training did not complete. Go back to Step 5 and run all cells again. Make sure the last cell runs without error.

Port 5001 already in use

Open app.py in a text editor and find the last line. Change the port number from 5001 to 5002 or any available port.

librosa errors on Apple Silicon

Run the following command to install the required audio backend libraries.

pip install resampy soundfile

tensorflow import error

Try installing the CPU-only version of TensorFlow.

pip install tensorflow-cpu

Slow or broken browser recording

Use Google Chrome or Microsoft Edge. Safari has known limitations with the MediaRecorder API and may not record audio correctly.

Wrong predictions after modifying config

Do not edit config.pkl manually. If you changed any preprocessing parameters, delete config.pkl and retrain the model completely from Step 4.