一种基于私有区块链的信息保护预测模型研究
发布时间：2018-08-07 13:15
【摘要】：[目的/意义]预测建模是数据挖掘的基础任务之一,当前基于信息保护的预测模型大多建立在一个中心化的架构之下,因而不可避免地存在一些安全性和鲁棒性漏洞。因此尝试一种新的、去中心化的预测建模方法,能够兼具敏感信息保护和数据处理能力,无论对跨机构互操作还是对国家层面的信息安全与共享服务,均有重要的意义。[方法/过程]文章提出了一个新的预测模型架构——模型链,将基于信息保护的在线机器学习与私有区块链网络技术相互整合,应用交易元数据传递局部模型,并提出一种新的信息证明算法以确定在线学习进程的执行顺序。[结果/结论]在模型链中,每个参与的站点都将有助于模型参数估计,但无需透露己方的任何信息(即只有模型参数而没有用户数据在机构间进行交换)。基于区块链技术的预测建模能够有效提高机构间互操作性,同时规避了敏感信息泄露、数据处理需要中央服务器引导而产生的种种系统性、政策性风险。该方法将有助于支持全国范围内的信息共享服务的路线图设计。[局限]模型链的技术实施还有待结合具体的网络环境进行评估,这也是开展后续研究的目标。
[Abstract]:[objective / significance] Prediction modeling is one of the basic tasks of data mining. Currently, most of the prediction models based on information protection are based on a centralized framework, so there are inevitably some security and robustness vulnerabilities. Therefore, it is of great significance to try a new, decentralized predictive modeling method with both sensitive information protection and data processing capabilities, both for cross-agency interoperability and for information security and sharing services at the national level. [method / process] this paper proposes a new predictive model architecture, model chain, which integrates online machine learning based on information protection with private block chain network technology, and applies transaction metadata to transfer local model. A new information proof algorithm is proposed to determine the execution order of the online learning process. [results / conclusions] in the model chain, each participating site will contribute to the estimation of model parameters, but there is no need to disclose any information about the model (that is, only model parameters and no user data are exchanged between agencies). The prediction modeling based on block chain technology can effectively improve the interoperability between institutions and avoid the systematic and policy risks caused by the disclosure of sensitive information and the need for the central server to guide data processing. This approach will help support the roadmap design of information-sharing services across the country. The technical implementation of the model chain needs to be evaluated in the light of the specific network environment, which is also the goal of further research.
【作者单位】： 武汉大学信息管理学院;武汉大学计算机学院;
【基金】：国家自然科学基金项目“融合情境的移动阅读推荐系统研究”(项目编号:71373192),国家自然科学基金项目“面向移动终端的情境隐私保护理论与方法研究”(项目编号:61572380)和国家自然科学基金面上项目“无线Mesh网络中跨层安全关键技术研究”(项目编号:61272451)的部分成果
【分类号】：G203
本文编号：2170136
[Abstract]:[objective / significance] Prediction modeling is one of the basic tasks of data mining. Currently, most of the prediction models based on information protection are based on a centralized framework, so there are inevitably some security and robustness vulnerabilities. Therefore, it is of great significance to try a new, decentralized predictive modeling method with both sensitive information protection and data processing capabilities, both for cross-agency interoperability and for information security and sharing services at the national level. [method / process] this paper proposes a new predictive model architecture, model chain, which integrates online machine learning based on information protection with private block chain network technology, and applies transaction metadata to transfer local model. A new information proof algorithm is proposed to determine the execution order of the online learning process. [results / conclusions] in the model chain, each participating site will contribute to the estimation of model parameters, but there is no need to disclose any information about the model (that is, only model parameters and no user data are exchanged between agencies). The prediction modeling based on block chain technology can effectively improve the interoperability between institutions and avoid the systematic and policy risks caused by the disclosure of sensitive information and the need for the central server to guide data processing. This approach will help support the roadmap design of information-sharing services across the country. The technical implementation of the model chain needs to be evaluated in the light of the specific network environment, which is also the goal of further research.
【作者单位】： 武汉大学信息管理学院;武汉大学计算机学院;
【基金】：国家自然科学基金项目“融合情境的移动阅读推荐系统研究”(项目编号:71373192),国家自然科学基金项目“面向移动终端的情境隐私保护理论与方法研究”(项目编号:61572380)和国家自然科学基金面上项目“无线Mesh网络中跨层安全关键技术研究”(项目编号:61272451)的部分成果
【分类号】：G203
【相似文献】
相关期刊论文 前3条
1 邓赶生;;个人信息保护之立法理念研究——在信息保护与信息自由之间[J];法制博览(中旬刊);2013年01期
2 邢国春;陈星霖;;一种新型网络入侵检测系统对信息保护的预取模型[J];情报科学;2012年06期
3 ;数字[J];传媒;2013年01期
相关重要报纸文章 前2条
1 刘德良 北京师范大学教授 亚太网络法律研究中心主任;让网络信息保护更具可操作性[N];中国纪检监察报;2013年
2 石鹰;信息保护构建“公民本位”社会[N];经济视点报;2008年
，本文编号：2170136
本文链接：https://www.wllwen.com/xinwenchuanbolunwen/2170136.html
