Gaze3P Dataset

Paper:

'Gaze3P: Gaze-Based Prediction of User-Perceived Privacy'

Description:

The Gaze3P artifact includes the full proposed dataset (stimuli, gaze data, and participants' demographics), the data collection software (for future data augmentation and reproducibility), the data processing code (to convert the eye tracking data into readable formats for further processing).

Security/Privacy Issues and Ethical Concerns:

Compliance with the privacy and ethics guidelines: The data was collected and processed according to the standards, guidelines, and approval of the ethical committee of the authors’ institution, with participants’ consent, remuneration, and pseudoanonymisation procedures. In particular, the privacy and ethical guidelines of the Menlo Report were satisfied.

Environment:

We provide a full description of how to use the data collection software and the raw gaze conversion code in a README file. We recommend using Windows 7 or later for compatibility with the EyeLink software.

Accessibility:

Artifacts uploaded to Zenodo and  Dropbox:

DOI: https://doi.org/10.5281/zenodo.17104154

Notes on Reusability:

The dataset can be used to infer further insights from the data. The data collection software can be used for future data augmentation and reproducibility.

To facilitate an effective starting point for working with the Gaze3P dataset, researchers are encouraged to first explore the dataset’s structure and properties, and then train baseline models following the procedures outlined here. (Note: This is particularly useful for researchers with expertise in differential privacy (DP), as it provides a practical introduction to standard ML workflows.)