网络广告中防范欺诈点击技术的研究

发布时间：2018-02-09 13:44

本文关键词： 网络广告点击欺诈用户行为异常检测博弈论　出处：《兰州交通大学》2014年硕士论文　论文类型：学位论文

【摘要】：随着科技的发展和媒体的进步，传统的广告经营模式已经满足不了发展的趋势，网络作为新型媒介平台，占据了市场比较重要的位置。由于网络信息的便捷性、及时性，网络平台也得到社会中各个商业领域的普遍应用。许多企业开始使用网络广告的途径扩展自己产品的知名度，使得广告费用支付方式被密切关注，其中按点击付费广告是当前互联网界操作简单、易于管理且备受用户青睐的广告计价方式。按点击付费模式带来计费便利的同时也出现了恶意的欺诈点击行为，此类点击的出现和泛滥，严重的威胁了网络市场的合理稳健发展。因此，如何预防及检测网络广告中点击欺诈的现象显得至关重要。网络广告中的点击欺诈出现的主要原因是市场中存在的牟取利益的非法竞争。目前点击欺诈防范方法，主要以预防与检测两个角度，从用户点击行为中点击流数据特性出发，结合防范策略进行研究。本文从已有的点击欺诈的识别技术方法出发，采用基于密度的异常检测算法对网络中用户行为分析，引入博弈控制思想，做出最后决策来构建模型。论文研究内容具体如下： (1)介绍了按点击付费模式的付费机制及特点，点击欺诈现象伴随出现的缘由。按点击付费模式与其他模式的区别在于对网络广告感兴趣的客户每点击一次产生一次付费。而市场中的恶意竞争者正是利用此付费机制中特点，进行恶意点击来谋取利益。 (2)通过研究当前存在的几种欺诈点击识别技术及原理，采用基于密度的异常检测算法进行欺诈行为的分类检测。 (3)在检测欺诈点击过程中，，引入博弈论机制，提出一种基于用户行为防范欺诈点击模型。 (4)本文使用真实的点击流数据信息作为实验数据，采用精度、召回率和准确率来衡量基于用户行为的防范欺诈点击模型的效果。实验表明，该模型具有较高的欺诈点击行为的识别能力，从而有效监测欺诈点击行为。
[Abstract]:With the development of science and technology and the progress of media, the traditional advertising management mode can not meet the trend of development. As a new media platform, the network occupies an important position in the market. The network platform has also been widely used in various commercial fields in society. Many enterprises begin to use the way of online advertising to expand the popularity of their products, so that the way of paying for advertising expenses is paid close attention to. Pay-per-click advertising is a simple, easy-to-manage and popular method of advertising pricing in the current Internet world. The click-per-click model brings convenience to billing, and malicious fraudulent click behavior also occurs at the same time. The emergence and proliferation of this kind of click seriously threaten the reasonable and steady development of the network market. Therefore, how to prevent and detect click fraud in online advertising is very important. The main reason of click fraud in network advertisement is the illegal competition for profit making in the market. At present, the methods of preventing click fraud mainly from the perspective of prevention and detection, proceed from the characteristics of click-stream data in user's click-through behavior. Based on the existing methods of click-fraud identification, the density based anomaly detection algorithm is used to analyze the behavior of users in the network, and the idea of game control is introduced. Make the final decision to build the model. This paper introduces the mechanism and characteristics of pay-per-click mode. Click fraud comes with it. The difference between the click-per-click model and other models is that customers who are interested in online advertising pay a fee every time they click. And malicious competitors in the market take advantage of this payment mechanism. Make a malicious click to make a profit. 2) by studying several existing techniques and principles of fraud click-through identification, the density based anomaly detection algorithm is used to detect fraud. In the process of detecting fraudulent click, the game theory mechanism is introduced, and a model of preventing fraud click based on user behavior is proposed. In this paper, the real click-stream data is used as the experimental data, and the accuracy, recall and accuracy are used to measure the effectiveness of the anti-fraud click-through model based on the user's behavior. The model has high recognition ability of fraud click behavior, so it can effectively monitor fraudulent click behavior.
【学位授予单位】：兰州交通大学
【学位级别】：硕士
【学位授予年份】：2014
【分类号】：TP393.08

【相似文献】