Network operators use classification systems to identify applications appearing in the traffic mixture in the internet. It is important to know the efficiency of the classification system in order to use the information provided by these systems for various management purposes and billing. Comparison of these systems can only be authoritative if it is based on realistic and up-to-date measurements, which in the same time do not contain sensitive data from network users. In this thesis, I describe a Java based framework that is capable of producing a traffic fulfilling the above mentioned requirements with emulating real user behaviour. The task of the thoroughly presented module of the framework is to aggregate the recorded traces when replaying user interactions in a way that they have the characteristics of the real trace, as well as providing the opportunity of verifying characteristics match with assembling different traffic attributes. After an introduction, I write about the implementation of the module and also give the performance evaluation of the system.