Gaussian Fitting of Multi-scale Traffic Properties for Discriminating IP Applications

摘要: Abstract —In the last years, there has been an increasingneed to accurately assign trafc its originating applicationor protocol. Several new protocols and services have appeared,such as VoIP or le sharing, creating additional identicationchallenges due their peculiar behaviors, such useof random ports associated other protocols. Thenumber variety of security vulnerabilities attacks thatare carried out over Internet also drastically increased inrecent years. Besides, privacy condentiality are growingconcerns for users: encryption is becoming widelyused and, therefore, access user payload more andmore difcult. Therefore, identication methodologies thatcan be accurate when applied different types andbe able operate in cyphered scenarios needed.In this paper, we present methodology thatrelies on a multiscale analysis ows, differentiatingthem based probability that characteristic multiscalebehavior estimators belong specic distributionswhose parameters inferred from ows real ap-plications. The classical concept ow was replacedby denition data stream , which consists all trafc(in upload download directions) local IP addressthat univocally identied by numeric identier. resultsachieved so far show proposed toaccurately classify licit identify some mostcommon attacks. approach can alsocircumvent most important drawbacks existingidentication methodologies, namely inability work understrict restriction scenarios.Keywords: Application identication, analysis,wavelets, illicit applications.