网络广告中的点击欺诈检测方法研究

发布时间：2018-05-05 08:26

本文选题：网络广告 + 点击欺诈　；参考：《哈尔滨工程大学》2016年硕士论文

【摘要】：随着互联网技术的飞速发展,网络服务已经渗透到了生活的方方面面,人们也逐渐形成了对网络信息的依赖,在网站中投放广告的方式已经成为目前网络中重要的宣传策略,同时也成为各个互联网公司重要的财政收入之一。目前国内外对于点击欺诈检测的相关研究,主要分为改进付费模式、验证码预防和数据分析检测三个方向。然而,对于点击欺诈的检测,仍存在广告商难以接受改变当前按点击付费的商业模式,验证码机制降低用户的体验好感,数据分析检测难以确定用户身份、数据庞大致使分析效率低等问题。本文针对以上这些问题,借鉴目前对点击欺诈检测的相关研究成果,在不改变计费模式的前提下,提出了一种基于分布式环境的点击欺诈检测模型。该点击欺诈检测模型主要包括数据采集、用户身份识别以及并行欺诈检测三个部分。数据采集模块通过客户端代码采集浏览器指纹、EverCookie存放数据来进行用户身份识别,同时采集点击相关数据进行并行欺诈检测;用户身份识别模块主要通过浏览器指纹识别和EverCookie投放相结合的方式对用户的身份进行识别;并行欺诈检测模块分为并行统计分析和并行可疑性分析,基于Hadoop分布式环境对用户点击数据进行并行分析检测。在本文的最后,通过实例数据对所研究的分析检测过程进行实验验证,结果表明本文提出的算法能够改进当前点击欺诈检测算法存在问题,在点击欺诈检测的效率和准确率方面有了明显提升,能够对网络广告中的点击欺诈行为进行可靠的检测。
[Abstract]:With the rapid development of Internet technology, network services have penetrated into all aspects of life, people have gradually formed a dependence on network information, the way to place advertisements in websites has become an important propaganda strategy in the current network. At the same time, it has also become one of the most important revenue for Internet companies. At present, the research on click fraud detection at home and abroad is mainly divided into three directions: improved payment mode, verification code prevention and data analysis and detection. However, for the detection of click fraud, advertisers still find it difficult to accept changing the current pay-per-click business model, the CAPTC-code mechanism reduces the user's good impression of the experience, and the data analysis detection is difficult to determine the identity of the user. The huge data result in low efficiency of analysis and so on. In order to solve these problems, this paper proposes a new click fraud detection model based on distributed environment without changing the billing model by referring to the current research results of click fraud detection. The click-fraud detection model includes three parts: data acquisition, user identification and parallel fraud detection. The data acquisition module uses the client code to collect the browser fingerprint and EverCookie to store the data to identify the user, and at the same time to collect and click the relevant data for parallel fraud detection. The user identification module mainly uses the combination of browser fingerprint identification and EverCookie to identify the identity of the user. The parallel fraud detection module is divided into parallel statistical analysis and parallel suspicious analysis. Parallel analysis and detection of user click-through data based on Hadoop distributed environment. At the end of this paper, the experimental results show that the proposed algorithm can improve the current click-fraud detection algorithm. The efficiency and accuracy of click fraud detection have been improved obviously, and the click fraud can be detected reliably in online advertising.
【学位授予单位】：哈尔滨工程大学
【学位级别】：硕士
【学位授予年份】：2016
【分类号】：TP393.092;TP301.6

【相似文献】