gstrag/TypeFormer

TypeFormer: Transformers for Mobile Keystroke Biometrics

This repository contains the source code of TypeFormer, a novel Transformer-based mobile keystroke verification system proposed in [1]. The model development and evaluation described is carried out over the Aalto Mobile Keystroke Database. The experimental protocol is the same as in [2], [3].

Model Development

The model was developed from the following sources:

• https://github.com/getalp/Lightweight-Transformer-Models-For-HAR-on-Mobile-Devices
• https://github.com/lucidrains/block-recurrent-transformer-pytorch
• https://github.com/BiDAlab/ExploringTransformers

Repo Description

In this repository, we provide:

• The source code for TypeFormer [1], and for the preliminary Transformer in [2] in model/Model.py and in model/Preliminary.py.
• The scripts to train (train.py) and evaluate (test.py) said models on the Aalto Mobile Keystroke Database.
• The scripts to train (KVC_train.py) and evaluate (KVC_test.py) said models on the Keystroke Verification Competition (KVC) .
• The script to plot DET curves (plot_DET.py).
• The script to plot loss and EER on the training and validation sets over the training epochs (read_log.py).
• The script to plot t-SNE representation (tSNE.py).
• Pretrained TypeFormer (pretrained/TypeFormer_pretrained.pt) and the preliminary Transformer (pretrained/preliminary_transformer_pretrained.pt).
• For the protocol adopted in the original papers [1, 2], adopted in the scripts train.py and test.py the preprocessed data on which we perform the experiments should be placed in data/. Get in touch with the authors to get access to the data.
• For the protocol adopted in the KVC, the data should be downloaded from CodaLab after joining the competition. The data should be placed in ../../databases/KVC_data/. This directory is specified in the variable configs.data_dir in the utils/KVC_config.py script and it can be modified. We recommend this approach over the original experimental protocol.

References

[1] Giuseppe Stragapede, Paula Delgado-Santos, Ruben Tolosana, Ruben Vera-Rodriguez, Richard Guest, and Aythami Morales, “TypeFormer: Transformers for Mobile Keystroke Biometrics”, arXiv:2207.07596, 2023.

[2] Giuseppe Stragapede, Paula Delgado-Santos, Ruben Tolosana, Ruben Vera-Rodriguez, Richard Guest, and Aythami Morales, “Mobile Keystroke Biometrics Using Transformers”, Proc. of the Int. Conf. on Face and Gesture Recognition (FG), 2023.

[3] Alejandro Acien, Aythami Morales, John V. Monaco, Ruben Vera-Rodriguez, and Julian Fierrez, "TypeNet: Deep learning keystroke biometrics." IEEE Transactions on Biometrics, Behavior, and Identity Science (TBIOM), 4.1 (2021): 57-70, 2021.

If you use any of the parts of this repo, please cite:

@article{stragapede2022typeformer,
  title={TypeFormer: Transformers for mobile keystroke biometrics},
  author={Stragapede, Giuseppe and Delgado-Santos, Paula and Tolosana, Ruben and Vera-Rodriguez, Ruben and Guest, Richard and Morales, Aythami},
  journal={arXiv preprint arXiv:2212.13075},
  year={2022}
}

@inproceedings{stragapede2023mobile,
  title={Mobile keystroke biometrics using transformers},
  author={Stragapede, Giuseppe and Delgado-Santos, Paula and Tolosana, Ruben and Vera-Rodriguez, Ruben and Guest, Richard and Morales, Aythami},
  booktitle={2023 IEEE 17th International Conference on Automatic Face and Gesture Recognition (FG)},
  pages={1--6},
  year={2023},
  organization={IEEE}
}