The widespread use of Large Language Models (LLMs) in society creates new information security challenges for developers, organizations, and end-users alike. LLMs are trained on large volumes of data, and their susceptibility to reveal the exact contents of the source training datasets poses security and safety risks. Although current alignment procedures restrict common risky behaviors, they do not completely prevent LLMs from leaking data. Prior work demonstrated that LLMs may be tricked into divulging training data by using out-of-distribution queries or adversarial techniques. In this paper, we demonstrate a simple, query-based decompositional method to extract news articles from two frontier LLMs. We use instruction decomposition techniques to incrementally extract fragments of training data. Out of 3723 New York Times articles, we extract at least one verbatim sentence from 73 articles, and over 20% of verbatim sentences from 6 articles. Our analysis demonstrates that this method successfully induces the LLM to generate texts that are reliable reproductions of news articles, meaning that they likely originate from the source training dataset. This method is simple, generalizable, and does not fine-tune or change the production model. If replicable at scale, this training data extraction methodology could expose new LLM security and safety vulnerabilities, including privacy risks and unauthorized data leaks. These implications require careful consideration from model development to its end-use.
Simple Yet Effective: Extracting Private Data Across Clients in Federated Fine-Tuning of Large Language Models
Harbin Institute of Technology（哈尔滨工业大学）；Kuaishou Technology（快手科技）；Meta AI；Monash University（墨尔本大学）；Fudan University（复旦大学）；Shanghai Academy of Artificial Intelligence for Science（上海人工智能科学研究院）
A Combination Model for Time Series Prediction using LSTM via Extracting Dynamic Features Based on Spatial Smoothing and Sequential General Variational Mode Decomposition
JiangXi University of Finances and Economics（江西财经大学）
Discriminant Dynamic Mode Decomposition for Labeled Spatio-Temporal Data Collections
RIKEN Center for Advanced Intelligence Project（理化学研究所先进智能项目中心）；University of Applied Sciences and Arts Western Switzerland（瑞士西部应用科学与艺术大学）；Graduate School of Informatics, Nagoya University（名古屋大学信息学研究科）；Graduate School of Informatics, Kyoto University（京都大学信息学研究科）；Institute of Mathematics for Industry, Kyushu University（九州大学产业数学研究所）
Variational decomposition autoencoding improves disentanglement of latent representations
Dept. of Biomedical Engineering, Khalifa University of Science and Technology（生物医学工程系，哈利法大学）；School of Electrical and Computer Engineering, Aristotle University of Thessaloniki（电气与计算机工程学院，亚里士多德大学）
Fast Learnings of Coupled Nonnegative Tensor Decomposition Using Optimal Gradient and Low-rank Approximation
Stem Cell Clinical Research Center, The First Affiliated Hospital of Dalian Medical University（大连医科大学附属第一医院干细胞临床研究中心）；Dalian Innovation Institute of Stem Cell and Precision Medicine（大连干细胞与精准医学创新研究院）；Faculty of Information Technology, University of Jyväskylä（于韦斯屈莱大学信息技术学院）；School of Biomedical Engineering, Faculty of Medicine, Dalian University of Technology（大连理工大学医学院生物医学工程学院）；Key Laboratory of Social Computing and Cognitive Intelligence (Dalian University of Technology), Ministry of Education（教育部社会计算与认知智能重点实验室）
Unsupervised EHR-based Phenotyping via Matrix and Tensor Decompositions
Simula Metropolitan Center for Digital Engineering（西姆拉大都会数字工程中心）；Oslo Metropolitan University（奥斯陆城市大学）；Swammerdam Institute for Life Sciences, University of Amsterdam（阿姆斯特丹大学斯瓦默丹生命科学研究所）
Characterization of Thermal Systems from Noisy and Low-resolution Measurements Using Dynamic Mode Decomposition
Universidade Estadual Paulista — UNESP（圣保罗州立大学）；Laboratório Nacional de Computação Científica — LNCC（国家科学计算实验室）；Universidade do Estado do Rio de Janeiro — UERJ（里约热内卢州立大学）
