jaschadub/VectorSmuggle

Testing platform for covert data exfiltration techniques where sensitive documents are embedded into vector representations and tunneled out under the guise of legitimate RAG operations — bypassing traditional security controls and evading detection through semantic obfuscation.

39
/ 100
Emerging

This project helps security researchers and defenders understand how sensitive information can be secretly hidden and extracted from AI systems, especially those using Retrieval-Augmented Generation (RAG). It takes various document formats as input, embeds hidden data into their vector representations, and then demonstrates how to query and reconstruct this data, effectively bypassing standard security measures. Security professionals, such as red teamers, security architects, and AI/ML security engineers, would use this to identify and mitigate such vulnerabilities.

Use this if you are an AI/ML security professional needing to test and understand how covert data exfiltration works in RAG-based systems.

Not ideal if you are looking for a general-purpose data encryption tool or a standard RAG system for legitimate data retrieval.

AI-security data-exfiltration RAG-security red-teaming information-security
No Package No Dependents
Maintenance 10 / 25
Adoption 8 / 25
Maturity 15 / 25
Community 6 / 25

How are scores calculated?

Stars

67

Forks

3

Language

Python

License

MIT

Last pushed

Feb 25, 2026

Commits (30d)

0

Get this data via API

curl "https://pt-edge.onrender.com/api/v1/quality/vector-db/jaschadub/VectorSmuggle"

Open to everyone — 100 requests/day, no key needed. Get a free key for 1,000/day.