Research On Estimation Of Domain Visits Via DNS Cache

As the use of Internet are gradually penetrating into all aspects of people’s life,obtaining reliable network demographics is becoming more and more important.Reliable network demographics can help people to develop more effective strategiesand to make better decisions. Traditional methods to obtain network demographicsare directly collecting client-side data from deployments of measurement agentsplaced inside edge networks, such as deploying toolbars in user’s browser. However,users are paying more and more attention on their privacy. They do not want you tocollect their personal information without their permissions. Moreover, since theclient-side nature of traditional methods, manipulation recipes, like click fraud, areabused, result in more and more difficult to identify the authenticity and integrity ofthose information collected by traditional methods.In order to strengthen the user privacy protection and guard against deception,we put forward a method to estimate domain visits based on DNS cache. The methodbelongs to the indirect measuring method. With this method, we can estimate thedomain visits without collecting any user’s information. This method does need todeploy any collection tools inside edge networks. Therefore, this method is lessinvasive and more robust against fraudulent behaviors. It is based on DNS cache.Due to the DNS service is the basis of many Internet services, this method is veryflexible. It can be applied to many fields.Secondly, we realize this measurement method, and point out the need toproperly handle some external factors in the implementation process. These factorsdirectly related to the accuracy of the measure results. For each external factors, weput forward the corresponding solutions.Finally, we use four domain names to test the method. These four domain nameare baidu, tecent, taobao and sina’s main domain name. We finally estimate visittraffic to these four domain names in Harbin Institute of Technology. Compared toactual number of clients, we found that the error of the measurement result is lessthan20%, in line with the forecast in advance. This indicates that this measurementmethod has high accuracy and great practical application value.