0:00 / 0:00

DEF CON 33 - Exploiting Shadow Data from AI Models and Embeddings - Patrick Walsh

Name: DEF CON 33 - Exploiting Shadow Data from AI Models and Embeddings - Patrick Walsh
Uploaded: 2025-10-10T21:41:27Z
Duration: 48 min 23 s
Channel: DEFCONConference
Description: Patrick Walsh demonstrates how to extract sensitive data from LLMs and RAG workflows using model and embedding inversion attacks.

DEFCONConference

419K subscribers • 61 videos on Hallway • Conference channel

YouTube

One of the most-watched uploads on DEF CON, a conference or university channel Hallway lists. Nobody has recommended it in the communities yet, so it ranks below community picks.
Presented by Patrick Walsh

Intermediate. Assumes working knowledge.

196K views1 year ago

Security AI and machine learning Exploits and vulnerabilities Transformers and LLMs

Patrick Walsh demonstrates how to extract sensitive data from LLMs and RAG workflows using model and embedding inversion attacks.

This talk explores the hidden risks in apps leveraging modern AI systems—especially those using large language models (LLMs) and retrieval-augmented generation (RAG) workflows. We demonstrate how sensitive data, such as personally identifiable information (PII) and social security numbers, can be extracted through real-world attacks. We’ll demonstrate model inversion attacks targeting fine-tuned models, and embedding inversion attacks on vector databases among others. The point is to show how PII scanning tools fail to recognize the rich data that lives in these systems and how much of privacy disaster these AI ecosystems really are.

Why it's here

A top upload from a listed channel

DEF CON is on Hallway's list of conference and university channels, and this upload is one of its most-watched talks or lectures. Nobody has recommended it in the communities yet, so it ranks below community picks.