This project investigates the security vulnerabilities of the ChatGPT plugin system, which allows developers to upload external programs to enhance user experience. While the ChatGPT model itself is secure, the security of its plugin ecosystem has not received sufficient attention.

With over 1,000 plugins covering various functionalities, this study aims to identify security risks associated with these plugins and evaluate the threats posed by external applications interacting with the ChatGPT model. The findings will highlight security issues and provide recommendations for improving the plugin system's security.