[Backlog]: Reverse Engineering Embeddings

[felipepenha]

Checklist

• Backlog entry requires creating new sandboxes.
• Backlog entry requires creating new exploitation code and/or tutorials.

CVE List

No response

Description

Summary

Exploitation that obtain text from a vector database.

GenAI Red Teaming Manual Reference

4.2.2.1 Embedding Inversion Attacks / A. Reverse Engineering Embeddings

Sandbox

Reuse sandboxes/RAG_local, which stores embedding vectors locally with chromadb.

Developer must check whether any change to sandboxes/RAG_local is necessary.

Exploitation

Obtain access to the sandbox vector database and the embedding model API.

The embedding model will be probed by coupling

https://github.com/ranfysvalle02/hacking-vectors

The learned model will, then, be used to invert vectors from the vector database.