Data Exfiltration via Markdown Images in Chatbots
Fatti validati
If a chatbot is manipulated through prompt injection, it can encode sensitive data in the link of an invisible image (``). When the chat reply is rendered, the browser loads the image and steals the data.
Spiegato in modo semplice
The secret postcard: The AI writes you a normal reply but sticks an invisible stamp on the back that sends your password to the attacker.
Rilevanza per la sicurezza e consiglio pratico
Chatbot UIs must strictly filter external image requests and HTML rendering.
Fonte primaria / riferimento
- Embrace The Red Cyber Research