期刊
CANCER BIOMARKERS
卷 33, 期 2, 页码 185-198出版社
IOS PRESS
DOI: 10.3233/CBM-210306
关键词
Privacy; privacy-preserving training; deep learning; natural language processing; cancer epidemiology; artificial intelligence
类别
资金
- National Nuclear Security Administration [17-SC-20-SC]
- US Department of Energy (DOE) Office of Science [17-SC-20-SC]
- Joint Design of Advanced Computing Solutions for Cancer (JDACS4C) program by DOE
- NCI of the National Institutes of Health
- DOE [DE-AC02-06-CH11357, DE-AC52-07NA27344, DE-AC5206NA25396, DE-AC05-00OR22725]
- Laboratory Directed Research and Development (LDRD) program of Oak Ridge National Laboratory, under LDRD project [9831]
- California Department of Public Health [103885]
- Centers for Disease Control and Prevention's (CDC) National Program of Cancer Registries [5NU58DP006344]
- National Cancer Institute's Surveillance, Epidemiology and End Results Program [HHSN261201800032I, HHSN261201800015I, HHSN261201800009I]
- NCI Surveillance, Epidemiology and End Results (SEER) Program [HHSN261201800013I]
- CDC National Program of Cancer Registries (NPCR) [U58DP00003907]
- Commonwealth of Kentucky
- NCI [HHSN26120180000 7I, HHSN26120 1300021I]
- Surveillance, Epidemiology and End Results (SEER) Program [HHSN26120180000 7I, HHSN26120 1300021I]
- CDC's National Program of Cancer Registries (NPCR) [NU58DP006332-02-00, NU58DP006279-02-00]
- State of Louisiana
- State of New Jersey
- Rutgers Cancer Institute of New Jersey
- National Cancer Institute's Surveillance, Epidemiology and End Results (SEER) Program [HHSN26120180001 4I, HHSN26100001]
- National Cancer Institute's SEER Program [HHSN261291800004I, HHSN261201800016I]
- Fred Hutchinson Cancer Research Center
- US Centers for Disease Control and Prevention's National Program of Cancer Registries [NU58DP0063 200]
- University of Utah
- Huntsman Cancer Foundation
- DOE Office of Science [DE-AC05-00OR22725]
- US Department of Energy (DOE) [DE-AC05-00OR22725]
- DOE
With the advancement of AI and machine learning in biomedical informatics, the concern for data security and privacy has become crucial. This study aims to quantify the privacy vulnerability of deep learning models for information extraction from medical textural contents, and propose ways to secure patients' information. Results show that the proposed vocabulary selection methods reduce privacy vulnerability while maintaining clinical task performance.
BACKGROUND: With the use of artificial intelligence and machine learning techniques for biomedical informatics, security and privacy concerns over the data and subject identities have also become an important issue and essential research topic. Without intentional safeguards, machine learning models may find patterns and features to improve task performance that are associated with private personal information. OBJECTIVE: The privacy vulnerability of deep learning models for information extraction from medical textural contents needs to be quantified since the models are exposed to private health information and personally identifiable information. The objective of the study is to quantify the privacy vulnerability of the deep learning models for natural language processing and explore a proper way of securing patients' information to mitigate confidentiality breaches. METHODS: The target model is the multitask convolutional neural network for information extraction from cancer pathology reports, where the data for training the model are from multiple state population-based cancer registries. This study proposes the following schemes to collect vocabularies from the cancer pathology reports; (a) words appearing in multiple registries, and (b) words that have higher mutual information. We performed membership inference attacks on the models in high-performance computing environments. RESULTS: The comparison outcomes suggest that the proposed vocabulary selection methods resulted in lower privacy vulnerability while maintaining the same level of clinical task performance.
作者
我是这篇论文的作者
点击您的名字以认领此论文并将其添加到您的个人资料中。
推荐
暂无数据