Tensor Trust
Tensor Trust is a web-based tool designed to explore prompt injection attacks while collecting valuable data to enhance the robustness of machine learning models. It involves a game format that allows researchers and developers to engage in interactive scenarios, providing insights into adversarial machine learning challenges.
Key Features:
- Interactive Gameplay: Engage with a unique prompt injection game that simulates real-world scenarios.
- Data Collection: Compile data crucial for understanding and improving the resilience of ML models against adversarial attacks.
- Django Framework: Built using Django, which ensures a robust and scalable development environment.
- Deployment: Easily deployable on Google Cloud Platform, leveraging various GCP services for better performance.
Benefits:
- Robustness Research: A valuable resource for researchers focusing on the security and robustness of machine learning applications.
- Collaborative Environment: Opens avenues for collaboration within the AI research community.
- Hands-on Learning: Provides a space for users to learn about prompt injection in an engaging manner.
