【作者】 赵丹；

【导师】 胡东辉；

【作者基本信息】 合肥工业大学 ， 信息安全， 2019， 硕士

【摘要】 计算机网络和大数据技术的快速发展为数据在世界范围内的迅速传播开辟了便捷的途径。如今,音频、视频、图像和文字等各种形式的数据在网络上的发布比比皆是。数据库作为数据存储的主要载体,频发的数据安全事件更是让人们感受到数据库版权和数据隐私的重要性。虽然目前已经有不少工作分别通过水印和差分隐私保护等方式对数据库进行版权保护和隐私保护,但是这些工作还存在以下不足:(1)目前的鲁棒性可逆数据库水印技术在处理整数型数据时导致的数据失真较大,不符合大数据共享环境下数据挖掘等方法对添加水印后数据库可用性需求;(2)在目前日趋流行的大数据交易的场景下,需要同时考虑数据库的版权跟踪和隐私保护问题,然而目前缺乏同时支持版权保护和隐私保护的数据库水印技术。针对以上这些问题,本文从以下方面展开研究:首先,本文设计了一种有失真限制的鲁棒性可逆数据库水印方法。根据直方图平移造成的数据失真较小的特点,我们设计了一种新型直方图平移可逆水印方法,然后利用遗传算法对水印方法进行优化,根据水印的嵌入容量和水印造成数据失真之间的矛盾,设计了遗传算法的适应度函数以期取得水印嵌入容量和数据失真之间的平衡。其次,本文提出并设计了一种基于本地差分隐私的数据库水印的方法。该水印方法以本地差分隐私技术为基础,根据本地差分隐私的不同实现机制,分别设计了基于拉普拉斯机制和基于随机响应机制的数据库水印方法。数据库水印方法依据本地差分隐私造成的较大的数据冗余来嵌入水印,该设计在减小数据失真,提高数据质量的同时也实现了版权保护的功能。最后,本文通过大量的实验证实了可逆数据库水印的有效性,以及数据库在遭受到各种恶意攻击时水印的鲁棒性;通过理论证明和实验分析验证了所设计的基于本地差分隐私的数据库水印方法的有效性和可行性,以及水印对隐私保护的影响,通过实验对比验证了数据的可用性以及两种基于本地差分隐私的数据库水印方法的优缺点。更多还原

【Abstract】 The rapid development of computer network and big data technology provide much convenience for information sharing.Nowadays,various forms of data,such as audio,video,image and text,are published on the Internet.As the main carrier of data storage,database is facing urgent requirments of copyright proteciotn and privacy protection with the data security incidents being reported frequently around the world.Large number of research works have been proposed to protect the copyright and privacy of the database through watermarking and differential privacy,respectively;however,these works have the following shortcomings:(1)state-of-the-art database watermarking schemes introduce large data distortion when dealing with integer data,which cannot satify the utility requirement of data mining aolgorithms on the watermarked database in the circumtances of bigdata sharing;(2)in the scenario of growing popularity of big data transactions,database sharing needs to consider both the copyright protection and data privacy protection,while state-of-the-art database watermarking schemes cannot provide both copyright protection and data privacy protection at the same time.Aiming at these problems,this dissertation conducts the following research:Firstly,this dissertation designs a robust reversible database watermarking method with distortion control.According to the characteristics of small data distortion caused by histogram shifting,we design a novel histogram shifting reversible watermarking method.The genetic algorithm is used to optimize the watermarking method.According to the contradiction between the watermarking capacity and data distortion caused by watermarks,we design a fitness function of genetic algorithm so as to achieve the balance between watermarking capacity and data distortion.Secondly,this dissertation proposes and designs database watermarking methods based on local differential privacy.According to the different implementation mechanisms of local differential privacy,database watermarking methods based on Laplacian mechanism and randomized response mechanism are proposed respectively.The database watermarking methods utilize the large data redundancy caused by local differential privacy to embed the watermark,which reduce data distortion and improve data utility at the same time.Finally,this dissertation proves the effectiveness of reversible database watermarking method and the robustness of the watermark through a large number of experiments;and this dissertation proves the effectiveness and feasibility of the local differential privacy-based database watermarking methods,analyzes the impact of watermarking on privacy protection through theoretical proof and experimental results,as well as verifies the data utility and the advantages and disadvantages of each proposed watermarking method by comparative experiments.更多还原