近年来，随着Internet的不断发展，网络成为了广告炙手可热的载体，网络广告的迅猛发展为广告主和互联网站带来了无限的商机。但是，由于国际流行广告CPC(Cost Per Click)计价方式为不法分子提供了获取非法利益的机会，所以点击欺骗现象变得日益猖獗。点击欺骗存在于在线搜索广告,按点击付费模式中。网络在线广告的点击欺骗方式不仅复杂而且难以检测，其检测技术成为了近些年因特网技术中新兴的研究课题。 论文首先总结了目前国内外存在的广告点击计价方式，并针对CPC计价方式着重介绍了存在的几种点击欺骗模式。对于存在的点击欺骗现象，现有的点击欺骗检测方法主要是以相对静态的身份和角色为基础，不能很好的说明未来用户行为的趋势，缺乏必要的信任预测，并且有些检测方法以严重牺牲网络用户上网体验为代价。通过对这些检测方法的分析和比较，作者提出了基于用户行为广告点击欺骗检测方法。该方法主要是研究网络用户点击行为，然后再根据用户的行为进行动态控制决策。在该方法实现过程中，作者首先利用一定技术手段获取客户端访问数据，然后选择该检测方法所需要的数据项建立点击流数据仓库，接着作者利用数据挖掘中贝叶斯网络分类方法预测每次客户点击的合法性等级概率，最后将客户点击预测结果和博弈控制相结合对双方的支付矩阵进行分析，计算出了基于用户不同行为属性的混合纳什均衡策略，最后判定用户每次点击行为的真实合法性。 在论文的最后，作者结合实例检验了所提出的基于用户行为广告点击欺骗检测方法，并取得了良好的效果。本文的结果对于量化分析用户广告点击行为合法性具有重要的理论意义，因此在实际网络应用中也具有重要的指导作用．

In recent years, with the Continuous development of internet, the network has become a hot advertising carrier; the rapid development of Network advertisement has brought infinite business opportunities for advertisers and Websites. However, because valuation ways of international popular advertising CPC(Cost Per Click)have provided access to benefits for outlaws, so Click fraud phenomenon has become increasingly rampant. Click fraud exists in the online search advertisement, Click payment pattern. Click fraud ways of the network Online advertisement not only complex but also hard to detect, the detection technology has become the emerging research subject in the Internet technology in recent years. Firstly, the paper summarized the current domestic and foreign existence advertisement click valuation ways, and introduced the existence several kinds of Click fraud patterns emphatically, which aiming at the CPC valuation way. For existing click fraud phenomenon, existing click fraud detection method is mainly by relative static identity and role as the foundation, not very good note the future trend of user behavior, lack of necessary trust, and some detection methods take the serious sacrifice network users Internet experience as a cost. Through the analysis and comparison of these detection methods, the author proposed advertisement click fraud detection method based on the user behavior. This method mainly studies the network user click behavior, and then according to user’s behavior conducts dynamic control decision-making. in the process of this method, the author first uses certain techniques to gain clients to access data, and then select needed data items of the detection method to establish click stream data warehouse, And then the author use data mining the Bayesian network classification method to forecast each time customer click valid rank probability, and finally combines the customers click forecasting result and the gambling control to analyze the bilateral payoff matrix, calculated attribute mixed the Nash balanced strategy based on the user different behavior, finally judged users each click behavior real validity . At the end of this paper, combining the examples, the author checked out the advertisement click fraud detection method based on the user behavior, and achieved good effect. The result of this paper has an important theoretical significance on quantizing analysis legality of the user advertising click behavior. Therefore, also has an important guiding role in the actual network applications