TrustAIRLab
Published January 28, 2025 | Version 0.1
Dataset Restricted

On the Proactive Generation of Unsafe Images From Text-To-Image Models Using Benign Prompts

Authors/Creators

  • 1. ROR icon Helmholtz Center for Information Security
  • 2. ROR icon Netflix (United States)
  • 3. ROR icon NetApp (United States)

Contributors

Project leader:

Project member (2):

  • 1. ROR icon Helmholtz Center for Information Security
  • 2. ROR icon Netflix (United States)
  • 3. ROR icon NetApp (United States)

Description

Due to ethical considerations, the datasets for our paper, including the targeted hateful memes dataset (`data/toxic_images/`) and the poisoning datasets (`data/unsafe/`), are hosted on here with the request-access feature enabled.

Files

Restricted

The record is publicly accessible, but files are restricted. <a href="https://zenodo.org/account/settings/login?next=https://zenodo.org/records/14754526">Log in</a> to check if you have access.

Additional details

Software

Repository URL
https://github.com/TrustAIRLab/proactive_unsafe_generation
Programming language
Python
Development Status
Active