Fairness in Oculomotoric Biometric Identification

This repo provides the code for reproducing the experiments in Fairness in Oculomotoric Biometric Identification.

In the paper, we investigate the fairness of a biometric system based on eye-tracking data with respect to gender, ethnicity, and age. The figure above shows the embedding evolution of the DeepEyedentificationLive subnets across ethnicities.

Reproduce the experiments

Download the data

You can download the publicly available data here: GazeBase, a large-scale, multi-stimulus, longitudinal eye movement dataset. The corresponding paper can be found here.

Clone this repository

You can clone this repository by either using

git clone git@github.com:aeye-lab/etra-fairness
cd etra-fairness

or

git clone https://github.com/aeye-lab/etra-fairness
cd etra-fairness

depending on your preferences and settings. Afterward, change into the directory by using cd etra-fairness.

Install packages

Install all required python packages via:

pip install -r requirements.txt

Extract data

After moving the zipped gazebase data download in the first step into the repository, extract all files by executing:

python3 extract_gazebase_data.py

Then you can directly start using the DeepEyedentification network by adjusting and executing the following scripts for the experiment you want to investigate. A description of the CLI arguments is available via python3 deepEye_fairness_gazebase.py --help.

A list of all bash scripts used is below. Note: Running the experiments, especially on CPU, will take some time.

Pipeline DeepEyedentification

• run experiments:
  • run_random_sampling_all_settings.sh
  • run_experiment_age.sh
  • run_experiments_VD1_VD2.sh
  • run_experiments_RAN.sh
  • run_experiments_HSS.sh
  • run_experiments_FXS.sh
  • run_experiments_BLG.sh
  • run_experiments_TEX.sh
• create score dicts:
  • run_create_score_dicts.sh

Unfortunately the Lohr et al. takes a bit more work. After downloading the data you have to adjust the following two scripts:

• extract the eye movement events (fixation, saccades, PSO)
  • adjust the path in the scripts from A novel evaluation of two related and two independent algorithms for eye movement classification during reading. The paper can be found here.
• adjust the statistical feature extraction: Study of an Extensive Set of Eye Movement Features: Extraction Methods and Statistical Analysis. The paper can be found here. Note, you don't only have to adjust the script to match your data path (the data created by MNH) but also such that the algorithm only takes one eye movement event as input instead of the complete sequence.

Afterwards copy the data to lohr_feature_data/. You can execute all experiments with the bash scripts below. CLI options are available via python3 lohr_fairness_gazebase.py --help.

Pipeline Lohr et al.

• run experiments:
  • run_experiments_lohr.sh
  • run_experiments_lohr_adam_w.sh
• create score dicts:
  • run_create_score_dicts_lohr.sh

Calculate fairness metrics and visualize results

You can calculate the fairness metrics and visualize the results from your experiments with the notebooks:

• Fairness:
  • plot_results_deepEye.ipynb
  • plot_results_lohr.ipynb
• Visualize embeddings:
  • plot_t-sne_visualization_deepEye.ipynb
  • plot_t-sne_visualization_lohr.ipynb
• Eye movement similarities for different demographics:
  • inspect_differences.ipynb

Contribution

If you find any issues, please open an issue in the issue tracker.

If you want, you can also test your own oculomotoric biometric models substituting it within the piplines described above.

Cite our work

If you use our code for your research, please consider citing our paper:

@inproceedings{10.1145/3517031.3529633,
author = {Prasse, Paul and Reich, David Robert and Makowski, Silvia and J\"{a}ger, Lena A. and Scheffer, Tobias},
title = {Fairness in Oculomotoric Biometric Identification},
year = {2022},
isbn = {9781450392525},
publisher = {Association for Computing Machinery},
address = {New York, NY, USA},
doi = {10.1145/3517031.3529633},
booktitle = {2022 Symposium on Eye Tracking Research and Applications},
articleno = {22},
numpages = {8},
keywords = {fairness, neural networks, biometrics},
location = {Seattle, WA, USA},
series = {ETRA '22}
}