Scalable Extraction Of Training Data From (production) Language Models

Nasr Milad, Carlini Nicholas, Hayase Jonathan, Jagielski Matthew, Cooper A. Feder, Ippolito Daphne, Choquette-choo Christopher A., Wallace Eric, Tramèr Florian, Lee Katherine. Arxiv 2023

[Paper]    
GPT Model Architecture Reinforcement Learning Security Training Techniques

This paper studies extractable memorization: training data that an adversary can efficiently extract by querying a machine learning model without prior knowledge of the training dataset. We show an adversary can extract gigabytes of training data from open-source language models like Pythia or GPT-Neo, semi-open models like LLaMA or Falcon, and closed models like ChatGPT. Existing techniques from the literature suffice to attack unaligned models; in order to attack the aligned ChatGPT, we develop a new divergence attack that causes the model to diverge from its chatbot-style generations and emit training data at a rate 150x higher than when behaving properly. Our methods show practical attacks can recover far more data than previously thought, and reveal that current alignment techniques do not eliminate memorization.

Similar Work