隐私保护数据挖掘*

隐私保护数据挖掘的目标是寻找一种数据集变换方法,使得敏感数据或敏感知识在实施数据挖掘的过程中不被发现。近年出现了大量相关算法,按照隐私保持技术可将它们分为基于启发式技术、基于安全多方技术和基于重构技术三种。结合目前研究的热点对关联规则和分类规则的隐私保护数据挖掘进行介绍,并给出算法的评估方法,最后提出了关联规则隐私保护数据挖掘未来研究工作的方向。     

同态加密隐私保护数据挖掘方法综述*

如何保护私有信息或敏感知识在数据挖掘过程中不被泄露,同时能得到较为准确的挖掘结果,是隐私保护中面临的重大挑战。近年来国内外学者对隐私保护数据挖掘（privacy-preserving data mining,PPDM）进行了大量研究,适时地对研究成果进行总结,能够明确研究方向。从分类挖掘、关联规则挖掘、聚类挖掘和安全多方计算等几个方面,总结了现有的基于同态加密技术的算法,分析了其基本原理和特点,并在此基础上指出了PPDM技术今后发展的方向。     

Reversible privacy preserving data mining: a combination of difference expansion and privacy preserving

Privacy Preserving Data Mining (PPDM) can prevent private data from disclosure in data mining. However, the current PPDM methods damaged the values of original data where knowledge from the mined data cannot be verified from the original data. In this paper, we combine the concept and technique based on the reversible data hiding to propose the reversible privacy preserving data mining scheme in order to solve the irrecoverable problem of PPDM. In the proposed privacy difference expansion (PDE) method, the original data is perturbed and embedded with a fragile watermark to accomplish privacy preserving and data integrity of mined data and to also recover the original data. Experimental tests are performed on classification accuracy, probabilistic information loss, and privacy disclosure risk used to evaluate the efficiency of PDE for privacy preserving and knowledge verification.     

数据挖掘隐私保护综述

随着社会信息化和电子商务与电子政务的不断发展,数据成为社会的重要资源,数据挖掘技术的应用逐渐深入。与此同时,隐私保护方面的问题已经成为数据挖掘研究的热点问题之一。本文介绍了数据挖掘隐私保护的发展现状,阐述了相关的概念、特征、分类和研究成果,并从数据扰动和多方安全计算两个方面介绍了数据挖掘隐私保护的相关技术,提出了未来的研究方向。     

Accurate and large-scale privacy-preserving data mining using the election paradigm

With the proliferation of the Web and ICT technologies there have been concerns about the handling and use of sensitive information by data mining systems. Recent research has focused on distributed environments where the participants in the system may also be mutually mistrustful. In this paper we discuss the design and security requirements for large-scale privacy-preserving data mining (PPDM) systems in a fully distributed setting, where each client possesses its own records of private data. To this end we argue in favor of using some well-known cryptographic primitives, borrowed from the literature on Internet elections. More specifically, our framework is based on the classical homomorphic election model, and particularly on an extension for supporting multi-candidate elections. We also review a recent scheme [Z. Yang, S. Zhong, R.N. Wright, Privacy-preserving classification of customer data without loss of accuracy, in: SDM’ 2005 SIAM International Conference on Data Mining, 2005] which was the first scheme that used the homomorphic encryption primitive for PPDM in the fully distributed setting. Finally, we show how our approach can be used as a building block to obtain Random Forests classification with enhanced prediction performance.     

Privacy-preserving data mining: A feature set partitioning approach

In privacy-preserving data mining (PPDM), a widely used method for achieving data mining goals while preserving privacy is based on k-anonymity. This method, which protects subject-specific sensitive data by anonymizing it before it is released for data mining, demands that every tuple in the released table should be indistinguishable from no fewer than k subjects. The most common approach for achieving compliance with k-anonymity is to replace certain values with less specific but semantically consistent values. In this paper we propose a different approach for achieving k-anonymity by partitioning the original dataset into several projections such that each one of them adheres to k-anonymity. Moreover, any attempt to rejoin the projections, results in a table that still complies with k-anonymity. A classifier is trained on each projection and subsequently, an unlabelled instance is classified by combining the classifications of all classifiers.Guided by classification accuracy and k-anonymity constraints, the proposed data mining privacy by decomposition (DMPD) algorithm uses a genetic algorithm to search for optimal feature set partitioning. Ten separate datasets were evaluated with DMPD in order to compare its classification performance with other k-anonymity-based methods. The results suggest that DMPD performs better than existing k-anonymity-based algorithms and there is no necessity for applying domain dependent knowledge. Using multiobjective optimization methods, we also examine the tradeoff between the two conflicting objectives in PPDM: privacy and predictive performance.     

数据库中的知识隐藏

伴随着数据共享、隐私保护、知识发现等多重需求而产生的PPDM(privacy preserving data mining),成为数据挖掘和信息安全领域近几年来的研究热点.PPDM中主要考虑两个层面的问题:一是敏感数据的隐藏与保护;二是数据中蕴涵的敏感知识的隐藏与保护(knowledge hiding in database,简称KHD).对目前的KHD技术进行分类和综述.首先介绍KHD产生的背景,然后着重讨论敏感关联规则隐藏技术和分类规则隐藏技术,接着探讨KHD方法的评估指标,最后归结出KHD后续研究的3个方向:数据修改技巧中基于目标距离的优化测度函数设计、数据重构技巧中的反向频繁项集挖掘以及基于数据抽样技巧的通用知识隐藏方法设计.     

一种基于格的隐私保护聚类数据挖掘方法

由于云计算的诸多优势,用户倾向于将数据挖掘和数据分析等业务外包到专业的云服务提供商,然而随之而来的是用户的隐私不能得到保证.目前,众多学者关注云环境下敏感数据存储的隐私保护,而隐私保护数据分析的相关研究还比较少.但是如果仅仅为了保护数据隐私,而不对大数据进行挖掘分析,大数据也就失去了其潜在的巨大价值.本文提出了一种云计算环境下基于格的隐私保护数据发布方法,利用格加密构建隐私数据的安全同态运算方法,并且在此基础上实现了支持隐私保护的云端密文数据聚类分析数据挖掘服务.为保护用户数据隐私,用户将数据加密之后发布到云服务提供商,云服务提供商利用基于格的同态加密算法实现隐私保护的k-means、隐私保护层次聚类以及隐私保护DBSCAN数据挖掘服务,但云服务提供商并不能直接访问用户数据破坏用户隐私.与现有的隐私数据发布方法相比,论文的隐私数据发布基于格的最接近向量困难问题（CVP）和最短向量困难问题（SVP）,具有很高的安全性.同时算法有效保持了密文数据间距离的精确性,与现有研究相比挖掘结果也具有更高的精确性和可用性.论文对方法的安全性进行了理论分析并设计实验对提出的隐私保护数据挖掘方法效率进行评估,实验结果表明本文提出的基于格的隐私保护数据挖掘算法与现有的方法相比具有更高的数据分析精确性和更高的计算效率.     

基于隐私保护的数据挖掘

基于隐私保护的数据挖掘(PPDM)的目标是在保护原始数据的情况下建立挖掘模型并得到理想的分析结果。该文从PPDM的总体需求出发,基于数据隐藏,将PPDM技术分为安全多方计算技术、匿名技术和数据转换技术。从准确性、隐私性和复杂性3个方面对PPDM技术进行了评估。     

利用PCA增强随机化隐私数据保护方法

基于随机化的数据扰乱及重构技术是数据挖掘中的隐私保护(Privacy-Preserving Data Mining,PPDM)领域中最重要的方法之一.但是,随机化难以消除由于属性变量本身相关性引起的数据泄漏.介绍了一种利用主成分分析(Principal Component Analysis,PCA)进行属性精简的增强随机化方法,降低了参与数据挖掘的属性数据间相关性,更好地保护了隐私数据.     

Using TF-IDF to hide sensitive itemsets

Data mining technology helps extract usable knowledge from large data sets. The process of data collection and data dissemination may, however, result in an inherent risk of privacy threats. Some sensitive or private information about individuals, businesses and organizations needs to be suppressed before it is shared or published. The privacy-preserving data mining (PPDM) has thus become an important issue in recent years. In this paper, we propose an algorithm called SIF-IDF for modifying original databases in order to hide sensitive itemsets. It is a greedy approach based on the concept borrowed from the Term Frequency and Inverse Document Frequency (TF-IDF) in text mining. The above concept is used to evaluate the similarity degrees between the items in transactions and the desired sensitive itemsets and then selects appropriate items in some transactions to hide. The proposed algorithm can easily make good trade-offs between privacy preserving and execution time. Experimental results also show the performance of the proposed approach.     

时空特性约束下的数据挖掘隐私保护方法

在隐私保护数据挖掘的研究中,隐私数据的时间特性以及空间特性是历来研究中常常被忽视的。将数据的安全级与时间性、空间性相结合,引入了数据安全级的时效性及空效性,然后采用层次概化方法进行数据隐私保护处理,并提出了基于时空特性的隐私保护关联规则挖掘算法。最后通过实验对算法的信息损失度、执行时间、算法效能等性能进行了分析和验证。     

PPDM中面向k-匿名的MI Loss评估模型

隐私保护数据挖掘（PPDM）利用匿名化等方法使数据所有者在不泄露隐私信息的前提下,安全发布在数据挖掘中有效可用的数据集。k-匿名算法作为PPDM研究使用最广泛的算法之一,具有计算开销低、数据形变小、能抵御链接攻击等优点,但是在一些k-匿名算法研究中使用的数据可用性评估模型的权重设置不合理,导致算法选择的最优匿名数据集在后续的分类问题中分类准确率较低。提出一种使用互信息计算权重的互信息损失（MI Loss）评估模型。互信息反映变量间的关联关系,MI Loss评估模型根据准标识符和标签之间的互信息计算权重,并通过Loss公式得到各个准标识符的信息损失,将加权后的准标识符信息损失的和作为数据集的信息损失,以弥补评估模型的缺陷。实验结果证明,运用MI Loss评估模型指导k-匿名算法能够明显降低匿名数据集在后续分类中的可用性丢失,相较于Loss模型和Entropy Loss模型,该模型分类准确率提升了0.73%~3.00%。     

面向数据特征的分布式数据挖掘研究

在众多提高数据挖掘效率的方法中,并行数据挖掘是一个从根本上解决该问题的有效途径.首先指出在数据挖掘过程中,不论采用顺序挖掘还是并行挖掘,都必须以数据挖掘的最终目的为前提,即尽可能多地发现数据中所含有的有用的知识,然后在此基础上提高数据挖掘的较率.在该想法基础上,提出了面向数据特征的数据划分过程,并进一步提出了加权式的并行数据挖掘基本方法.在这种数据挖掘过程中,可以得到相对于部分数据的知识,在很大程度上提高了数据挖掘的动态性能.     

A Framework for Evaluating Privacy Preserving Data Mining Algorithms<Superscript>*</Superscript>

Recently, a new class of data mining methods, known as privacy preserving data mining (PPDM) algorithms, has been developed by the research community working on security and knowledge discovery. The aim of these algorithms is the extraction of relevant knowledge from large amount of data, while protecting at the same time sensitive information. Several data mining techniques, incorporating privacy protection mechanisms, have been developed that allow one to hide sensitive itemsets or patterns, before the data mining process is executed. Privacy preserving classification methods, instead, prevent a miner from building a classifier which is able to predict sensitive data. Additionally, privacy preserving clustering techniques have been recently proposed, which distort sensitive numerical attributes, while preserving general features for clustering analysis. A crucial issue is to determine which ones among these privacy-preserving techniques better protect sensitive information. However, this is not the only criteria with respect to which these algorithms can be evaluated. It is also important to assess the quality of the data resulting from the modifications applied by each algorithm, as well as the performance of the algorithms. There is thus the need of identifying a comprehensive set of criteria with respect to which to assess the existing PPDM algorithms and determine which algorithm meets specific requirements. In this paper, we present a first evaluation framework for estimating and comparing different kinds of PPDM algorithms. Then, we apply our criteria to a specific set of algorithms and discuss the evaluation results we obtain. Finally, some considerations about future work and promising directions in the context of privacy preservation in data mining are discussed. ^*The work reported in this paper has been partially supported by the EU under the IST Project CODMINE and by the Sponsors of CERIAS. Editor:  Geoff Webb

---

Fuzzy data mining for time-series data

Time series analysis has always been an important and interesting research field due to its frequent appearance in different applications. In the past, many approaches based on regression, neural networks and other mathematical models were proposed to analyze the time series. In this paper, we attempt to use the data mining technique to analyze time series. Many previous studies on data mining have focused on handling binary-valued data. Time series data, however, are usually quantitative values. We thus extend our previous fuzzy mining approach for handling time-series data to find linguistic association rules. The proposed approach first uses a sliding window to generate continues subsequences from a given time series and then analyzes the fuzzy itemsets from these subsequences. Appropriate post-processing is then performed to remove redundant patterns. Experiments are also made to show the performance of the proposed mining algorithm. Since the final results are represented by linguistic rules, they will be friendlier to human than quantitative representation.     

动态数据挖掘研究

提出了一种新的数据挖掘形式--动态数据挖掘(DDM),寻求在不断更替产生的动态数据信息中找出能被应用的知识.给出动态数据挖掘的体系结构,并分析了动态数据挖掘实现过程,运用滑动窗口与动态数据窗口动态采集与处理动态新增数据,同时运用后续数据进行挖掘结果评价,用K标号法平滑地使用动态目标数据集进行数据挖掘,得出了一个动态数据挖掘测试算法.     

数据挖掘技术研究

数据挖掘是近年来数据库领域中出现的一个新兴研究热点。数据挖掘是对数据库数据的统计分析，其基础是人工智能。阐述了数据挖掘技术的背景，就数据挖掘的主要分析技术进行了研究和分类。介绍了目前数据挖掘常用的技术及数据挖掘过程，指出了数据挖掘技术未来的发展方向。     

Privacy-preserving collaborative data mining

Data collection is a necessary step in data mining process. Due to privacy reasons, collecting data from different parties becomes difficult. Privacy concerns may prevent the parties from directly sharing the data and some types of information about the data. How multiple parties collaboratively conduct data mining without breaching data privacy presents a challenge. The objective of this paper is to provide solutions for privacy-preserving collaborative data mining problems. In particular, we illustrate how to conduct privacy-preserving naive Bayesian classification which is one of the data mining tasks. To measure the privacy level for privacy- preserving schemes, we propose a definition of privacy and show that our solutions preserve data privacy.     

Emerging standards for data mining

This paper presents an overview of data mining, then discusses standards (both existing and proposed) that are relevant to data mining. This includes standards that affect several stages of a data mining project. Summaries of several emerging standards are given, as well as proposals that have the potential to change the way data mining tools are built.