Large language models (LLMs) show early signs of artificial general intelligence but struggle with hallucinations. One promising solution to mitigate these hallucinations is to store external knowledge as embeddings, aiding LLMs in retrieval-augmented generation. However, such a solution risks compromising privacy, as recent studies experimentally showed that the original text can be partially reconstructed from text embeddings by pre-trained language models. The significant advantage of LLMs over traditional pre-trained models may exacerbate these concerns. To this end, we investigate the effectiveness of reconstructing original knowledge and predicting entity attributes from these embeddings when LLMs are employed. Empirical findings indicate that LLMs significantly improve the accuracy of two evaluated tasks over those from pre-trained models, regardless of whether the texts are in-distribution or out-of-distribution. This underscores a heightened potential for LLMs to jeopardize user privacy, highlighting the negative consequences of their widespread use. We further discuss preliminary strategies to mitigate this risk.
KART: Parameterization of Privacy Leakage Scenarios from Pre-trained Language Models
The University of Tokyo(东京大学);The University of Tokyo Hospital(东京大学医院);Chiba University(千叶大学);Nara Institute of Science and Technology(奈良先端科学技术大学院大学)
对于安全共享预训练的语言模型,由于难以估计隐私泄露风险的上限,目前没有指导方针。一个问题是,之前的研究评估了不同真实世界隐私泄露场景和攻击方法的风险,这降低了研究结果的可移植性。为了解决这个问题,我们在通用参数化\textit{Knowledge, Anonymization, Resource, and Target}(KART)下表示复杂的现实世界隐私泄露场景。KART参数化有两个优点:(i)它澄清了每个实验中隐私泄露的定义;(ii)它提高了风险评估结果的可比性。我们表明,以前的研究可以简单地通过参数化的情况下与KART审查。我们还展示了隐私风险评估在不同的情况下,在相同的攻击方法,这表明KART有助于近似的风险上限下的特定攻击或场景。我们相信KART有助于整合过去和未来关于隐私风险的研究结果,并将为共享语言模型的标准做出贡献。
Privacy Challenges and Solutions in Retrieval-Augmented Generation-Enhanced LLMs for Healthcare Chatbots: A Review of Applications, Risks, and Future Directions
Centre for Smart Health, School of Nursing(智能健康中心、护理学院);The Hong Kong Polytechnic University(香港理工大学);Department of Electrical and Electronic Engineering(电子与电气工程系);Faculty of Health Sciences(健康科学学院);University of Maribor(马里博大学)
A Neighbourhood-Aware Differential Privacy Mechanism for Static Word Embeddings
University of Liverpool(利物浦大学);Amazon(亚马逊);National Institute of Informatics(国立信息学研究所);International Professional University of Technology in Tokyo(东京国际工科专门职大学)