Disputation: Desta Haileselassie Hagos
M.sc. Desta Haileselassie Hagos by the Department of Technology Systems will be defending his thesis
"Discovering the Dynamic Complexity of TCP Using Machine Learning and Deep Learning Techniques"
for the degree of Philosophiae Doctor.
The University of Oslo is closed. Desta`s PhD defence will therefore be fully digital and streamed directly using Zoom. The host of the session will moderate the technicalities while the chair of the defence will moderate the disputation.
Ex auditorio questions: the chair of the defence will invite the audience to ask ex auditorio questions either written or oral. This can be requested by clicking 'Participants -> Raise hand'.
"MultiPath TCP and RTP/RTCP protocols: motivations, mechanisms, deployment, and performances"
Download recording of the trial lecture. (Available two days in front of the disputation)
Conferral summery / Kreeringssammendrag
Doktorgradsarbeidet fokuserer hovedsaklig på monitorering av interne TCP tilstander i endenodene, basert på analyse av passive trafikkmålinger utført i en mellomliggende node i nettverket. Forskningen antas å være industrirelevant, fordi passiv trafikkmåling er en metode som i økende grad brukes av nettverksoperatører og Internet tjenestetilbydere for å analysere kommunikasjonsytelsen til nettverksbaserte applikasjoner og tjenester. Arbeidet foreslår ulike modeller og løsninger for å predikere TCP tilstander i endenoden, og presenterer eksperimenter som indikerer at prediksjonene gir relativt god nøyaktighet over ulike valideringsscenarier og over bruk av ulike TCP varianter.
Main research findings / Hovedfunn
The work presented in this dissertation aims at obtaining detailed knowledge about the end hosts by monitoring information of the packets that pass through the network, and by employing machine learning and deep learning-based techniques on the monitored network traffic. Since machine learning and deep learning methods are good at coping with complex tasks and massive amounts of data, they might play an important role to predict the TCP per-connection internal states. Understanding the dynamic complexity of the internal states of TCP is a fundamental challenge, and especially demanding due to the dynamics and complexity of modern networks. Even though this is the main objective of the dissertation, our work shows that related techniques can also be used to find other information about the hosts, such as their operating system or TCP implementation or in a security perspective classify if the host’s traffic is malicious or not.
The analyses of this dissertation focus mainly on TCP internal state monitoring from passive traffic measurements. We believe that our work will be useful for the industry since passive measurements are becoming increasingly useful for network operators and Internet Service Providers to evaluate the communication performance of applications and services running on their networks. Our experimental results indicate the effectiveness of the proposed prediction models with reasonably good accuracy across different validation scenarios and multiple TCP variants
Ida Elisabeth Rydning: firstname.lastname@example.org