OWASP Juice Shop is a popular, intentionally vulnerable web application that is often used by beginners and cybersecurity professionals to practice penetration testing.
This article focuses on one particular vulnerability found in web-based large language models (LLMs), which is listed as one of the top 10 vulnerabilities by OWASP 2025.
Prompt injection is an exploit where malicious actors manipulate LLMs by inputting prompts that appear legitimate but are designed to trigger unintended behaviour or bypass safety measures.
The author offers a tutorial on how to exploit the vulnerability in the OWASP Juice Shop, using the support chat function to receive a coupon code.
They also reference additional resources such as Gandalf and PortSwigger’s Web Security Academy, which are useful sites for learning about prompt injection and other LLM-based attacks.
The article serves as an encouraging introduction to the topic for those who are new to the concept.