Prompt Fuzzer by Prompt Security is an open-source tool designed for vulnerability assessment and hardening of Generative AI (GenAI) applications. It allows developers to evaluate and enhance the resilience and safety of their GenAI systems by simulating Large Language Model (LLM) based attacks. The tool focuses on testing GenAI apps against risks such as prompt injections, system prompt leaks, jailbreaking, and harmful content elicitations. Prompt Fuzzer provides an interactive environment for developers to test their GenAI applications, allowing for real-time assessment and adjustments. It performs dynamic LLM-based attacks and provides a security score along with an evaluation based on the test outcomes. Users can iteratively adjust chat format settings and modify the system prompt to improve safety and security scores. The tool is available on GitHub for public use and integration into development workflows.
