The Secret Sharer: Evaluating Unintended Memorization in Neural

Research question

Do generative sequence models memorise rare secrets such as credit-card numbers?

Key findings

Canary "exposure" metric quantifies memorisation of inserted secrets - Memorisation appears early in training and is not simply overfitting - Differential privacy eliminates the effect

Sources

arxiv.orghttps://arxiv.org/abs/1802.08232

Contribute

Spot an error, or know a record we should add? Members can propose additions and submit corrections.

Submit research Report a correction or request removal

Related in Research Paper Library

SpliTEE: Improving LLM Inference on Trusted Hardware with Differentially Private GPU Outsourcing

User prompts provided to large language models (LLMs) may contain sensitive or private information that can be misused by remotely deployed models, such as through inadvertent memorization during retraining. One way…

Training-data extraction and memorisation2026

Scalable Extraction of Training Data from (Production) Language Models

The divergence attack on ChatGPT, extracting personal information and verbatim text, showing alignment does not protect against memorisation leaks.

Training-data extraction and memorisation2023

Scalable Extraction of Training Data from (Production) Language Models

Shows that aligned production models still leak training data, including through the divergence attack that made ChatGPT emit memorised text after repeating a word, and that alignment hides rather than removes…

Training-data extraction and memorisation2023