保护用户点击流数据的在线行为广告框架的研究与设计

发布时间：2018-10-19 19:01

【摘要】：近10年来，随着网络在全世界范围内迅猛发展，，互联网已成为我们日常生活中必不可少的一部分。相比传统的广告投放方式，网络不仅是广告行业中的重要组成部分并逐渐成为主要的广告投放渠道。其中，在线行为广告（OBA）以其准确的用户偏好定位效果赢得了许多广告投放人的青睐。在线行为广告有两个必要的概念——跟踪和定位。为了能够准确定位用户的偏好，广告网络商(如google，yahoo)要收集用户的网络点击流信息(用户访问过网页的URL历史)。在这个跟踪过程中，广告网络商并没有得到用户的允许。根据我们的调查，随着人们对安全隐私的逐步重视，人们对广告网络商的上述跟踪行为作出了质疑--他们并不想被广告网络商跟踪。因此，怎么在保护用户的隐私信息的同时保持OBA框架对用户的偏好定位，成为众多网络隐私安全领域学者所关注的问题。时至今日，国内外的学者提出了一些技术解决方案来阻止广告网络商跟踪用户的行为。但是这些解决方案比较激进，否定并摒弃了广告网络商对用户的偏好定位作用。而本文提出了一个可以让用户接受的在线行为广告框架（AcceptableTracking简称：AT）来解决传统在线行为广告框架的隐私泄漏问题。一方面，AT在很大程度上保留了广告网络商分析用户偏好定位的作用。另一方面，它保护了用户的点击流信息（本文定义的用户私密信息）。 AT保留了现有的在线行为广告的商业模型。它不需要用户在Web浏览器安装插件来达到阻止广告网络商跟踪用户的作用。因此，AT对于目前在线行为广告行业来说是一个可行的折中解决方案。本文描述了AT的结构设计及改进方案。通过实验，我们验证了AT基于网页特征流的用户定位算法的可行性。另一方面，我们对AT进行了性能分析，证明其对浏览器性能和网络流量造成的负担是可接受的。最后，我们将AT跟其他6个OBA隐私问题的解决方案进行了对比。从比较结果可以看出,AT具有以下优势。第一，可行性高。由于AT不需要各方作出巨大的改变，因此AT的可行性大大提高；第二，准确的广告定位。由于AT沿袭了广告网络的广告定位功能，所以AT保留了准确定位广告的优点；第三，AT引入的保护组织让在线行为广告的营收实现了透明化，让其营收过程变得可审计；第四，跟其他在浏览器分析用户偏好的方案相比，AT可以使用所有的点击欺骗检测方法。
[Abstract]:In the past 10 years, with the rapid development of the Internet in the world, the Internet has become an essential part of our daily life. Compared with the traditional advertising mode, the network is not only an important part of the advertising industry, but also gradually become the main advertising channel. Among them, online behavioral advertising (OBA) has won the favor of many advertisers because of its accurate user preference positioning effect. Online behavioral advertising has two essential concepts-tracking and positioning. In order to accurately locate users' preferences, advertisers (such as google,yahoo) need to collect web click stream information (URL history of users who have visited web pages). In this tracking process, the advertising network provider did not get the user's permission. According to our survey, as people pay more and more attention to security and privacy, people question this behavior of advertisers-they do not want to be followed by advertisers. Therefore, how to protect the privacy information of users while maintaining the preference of users in the OBA framework has become a concern of many scholars in the field of network privacy security. Up to now, scholars at home and abroad have put forward some technical solutions to prevent advertisers from following users' behavior. However, these solutions are more radical, negating and abandoning the role of advertising network users' preferences. In this paper, an online behavioral advertising framework (AcceptableTracking), which is acceptable to users, is proposed to solve the privacy disclosure problem of traditional online behavioral advertising frameworks. On the one hand, AT retains to a large extent the role of advertising network providers in analyzing user preferences. On the other hand, it protects the user's click-stream information (). AT, defined in this paper, retains the existing commercial model of online behavioral advertising). It does not require users to install plug-ins in Web browsers to prevent advertisers from tracking users. As a result, AT is a viable compromise for the online behavioral advertising industry. This paper describes the structure design and improvement scheme of AT. Through experiments, we verify the feasibility of AT user location algorithm based on web feature stream. On the other hand, we analyze the performance of AT and prove that its burden on browser performance and network traffic is acceptable. Finally, we compared AT with six other OBA privacy solutions. From the comparison results, we can see that AT has the following advantages. First, the feasibility is high. Because AT does not need to make great changes, the feasibility of AT is greatly improved; second, accurate advertising positioning. Because AT follows the ad location function of the advertising network, AT retains the advantages of accurate advertising; third, the protection organization introduced by AT makes the revenue from online behavioral advertising transparent and its revenue process auditable; fourth, AT can use all of the Click spoofing methods compared with other schemes that analyze user preferences in browsers.
【学位授予单位】：中山大学
【学位级别】：硕士
【学位授予年份】：2013
【分类号】：TP393.08

【相似文献】