Goal

Identify a person by their keystrokes timing.

Usage

Dependencies

• python3
• tensorflow 2
• pandas
• numpy
• keyboard

Download prompts dataset

bash download_food_reviews.sh

Collect data

sudo python3 typing_data_generator.py

You will be asked your name. This name will be used as a class in the classifier. It is assumed that all data points collected under a name are for the same typing style.

When collecting a data point, the user will be prompted with a random food review from Amazon. This can be used to bypass typer's block.

Train model

python3 train.py

Test inference

sudo python3 infer.py

You will be prompted with a food review, like in the data collection phase. After a single collected data point, the model will predict who you are and output the name.

Notes

Current model architecture

• one 1D conv layer
• one hidden layer
• softmax with num_classes