Abstract. Online advertising has been suffering serious click fraud problem. Fraudulent publishers can generate false clicks using malicious scripts embedded in their web pages. Even widely-used security techniques like iframe cannot prevent such attack. In this paper, we propose a framework and associated methodologies to automatically and quickly detect and filter false clicks generated by malicious scripts. We propose to create an impression-click identifier which is able to link corresponding impressions and clicks together with a predefined lifetime. The impression-click identifiers are stored in a special data structure and can be later validated upon a click is received. To catering the requirements of click fraud detection, the proposed framework provides the capability of automatically deleting the outdated identifiers and the identifiers that have been clicked. Besides, the framework has the nice features of constant-time inserting and querying, low false negative rate and lo...
Yanlin Peng, Linfeng Zhang, J. Morris Chang, Yong