Results for “context-poisoning”

Probe RAG applications for prompt injection via poisoned retrieved context and embedding manipulation.

Exploit web cache mechanisms to serve malicious content to other users by poisoning cached responses through unkeyed headers and parameters during authorized security tests.