durdil/pwd_survey-main

0

stars

3

commits

Python

primary language

Nov 17, 2025

updated

README

pwd_survey

A web-based experimental platform developed for my bachelor's thesis to explore:

  • User-generated passwords
  • AI-generated password recommendations (GPT-API, DeepSeek, PassGPT)
  • Memorability and recall performance
  • Gaze-based behavioral data (eye tracking)

The experiment guides participants to create passwords, collects their inputs, generates AI-based password improvements, and stores all interaction data for later analysis.

Phases

Password Creation Phase

  • Users create passwords across multiple contexts (banking, shopping, email, social media).
  • The system generates password recommendations from:
    • GPT-API
    • DeepSeek API
    • PassGPT model
    • Randomized generator

Recall Phase

  • Users attempt to recall their chosen passwords.
  • Original options are displayed for comparison.

Memorability & Strength Evaluation Phase

  • Participants rate the passwords they used.
  • Stores rating data for later analysis.

Additional Implementations

Eye Tracking Features

  • Eye-tracking code is commented out for a full functioning experiment without a eye tracker
  • Note: To include eye tracking functionalities, uncomment the relevant parts marked with "TODO" in the app.py.

Analysis Scripts

  • The analysis/ directory includes some preprocessing and statistical scripts used for the thesis, one eye tracking preprocessing script has been left out since it was part of a private repository.
  • Note: paths such as PROJECT_ROOT must be adapted by users.

Deep Learning Models

  • Relevant keys are needed for full functionality using GPT-API, DeepSeek-API, PassGPT Model.
  • The application expects the following environment variables to be defined in your system environment:
  • if the keys are defined, uncomment the relevant parts marked with "TODO" in the utils.py.

Installation

The project uses a Conda environment defined in environment.yml.

conda env create -f environment.yml

conda activate pwd_survey

From the project root:

python app.py

This starts the web application locally.

Licence

All original code in this repository is licensed under the MIT License.

However, this project uses the PassGPT model, which is licensed under CC BY-NC 4.0 and is intended for research use only.

License details for PassGPT can be found here:
https://huggingface.co/javirandor/passgpt-10characters

Contributors

durdil

3 commits

durdil/pwd_survey-main

0

stars

3

commits

Python

primary language

Nov 17, 2025

updated

README

pwd_survey

A web-based experimental platform developed for my bachelor's thesis to explore:

  • User-generated passwords
  • AI-generated password recommendations (GPT-API, DeepSeek, PassGPT)
  • Memorability and recall performance
  • Gaze-based behavioral data (eye tracking)

The experiment guides participants to create passwords, collects their inputs, generates AI-based password improvements, and stores all interaction data for later analysis.

Phases

Password Creation Phase

  • Users create passwords across multiple contexts (banking, shopping, email, social media).
  • The system generates password recommendations from:
    • GPT-API
    • DeepSeek API
    • PassGPT model
    • Randomized generator

Recall Phase

  • Users attempt to recall their chosen passwords.
  • Original options are displayed for comparison.

Memorability & Strength Evaluation Phase

  • Participants rate the passwords they used.
  • Stores rating data for later analysis.

Additional Implementations

Eye Tracking Features

  • Eye-tracking code is commented out for a full functioning experiment without a eye tracker
  • Note: To include eye tracking functionalities, uncomment the relevant parts marked with "TODO" in the app.py.

Analysis Scripts

  • The analysis/ directory includes some preprocessing and statistical scripts used for the thesis, one eye tracking preprocessing script has been left out since it was part of a private repository.
  • Note: paths such as PROJECT_ROOT must be adapted by users.

Deep Learning Models

  • Relevant keys are needed for full functionality using GPT-API, DeepSeek-API, PassGPT Model.
  • The application expects the following environment variables to be defined in your system environment:
  • if the keys are defined, uncomment the relevant parts marked with "TODO" in the utils.py.

Installation

The project uses a Conda environment defined in environment.yml.

conda env create -f environment.yml

conda activate pwd_survey

From the project root:

python app.py

This starts the web application locally.

Licence

All original code in this repository is licensed under the MIT License.

However, this project uses the PassGPT model, which is licensed under CC BY-NC 4.0 and is intended for research use only.

License details for PassGPT can be found here:
https://huggingface.co/javirandor/passgpt-10characters

Contributors

durdil

3 commits

Languages

Python

75.7%

HTML

22.8%

CSS

1.5%