Procjena performansi strujanja igara zasnovanih na računalnom oblaku temeljena na analizi kriptiranog prometa uporabom metoda strojnog učenja

master's thesis

• Title (english): Streaming Performance Estimation for Cloud Gaming Based on the Analysis of Encrypted Network Traffic Using Machine Learning
• Author: Stipe Paradžik
• Mentor: Lea Skorin-Kapov (mentor)
• Committee member: Lea Skorin-Kapov (predsjednik povjerenstva)
• Committee member: Igor Sunday Pandžić (član povjerenstva)
• Committee member: Mirko Sužnjević (član povjerenstva)
• Granter: University of Zagreb; Faculty of Electrical Engineering and Computing; Zagreb
• Defense date and country: 2022-07-13, Croatia
• Scientific / art field, discipline and subdiscipline: TECHNICAL SCIENCES; Computing

Abstract

Zbog lakoće pristupa igre zasnovane na računalnom oblaku stječu sve veću popularnost među igračima, uz to izdavačima daju pristup široj publici i rješavaju problem piratizacije. Uz razvoj 5G mreža ovaj oblik igranja bi trebao biti još pristupačniji.
U ovome radu je analiziran utjecaj promjene parametara brzine kodiranja, broja okvira u sekundi i rezolucije na prostorne i vremenske video metrike, kao i analiza utjecaja tih istih parametara na mrežne podatke. Također su korištene tehnike strojnog učenja kako bi se predvidjeli rezolucija, brzine kodiranja i broj okvira u sekundi iz snimljenog mrežnog prometa za svaku odigranu sjednicu. Pri analizi utjecaja parametara na video metrike ustanovljeno je da promjena rezolucije ima velik utjecaj na prostorne, a promjena broja okvira u sekundi ima utjecaj na vremenske video metrike. Brzina kodiranja nije imala nikakvog utjecaja ni na prostorne ni na vremenske metrike. Također je uočeno da igre iz prvog i trećeg lica imaju visoke vremenske i niske prostorne metrike, dok igre s pogledom odozgo imaju niske vremenske i visoke prostorne video metrike. Analizom mrežnog prometa utvrdili smo da povećanjem brzine kodiranja, poveća se broj paketa kao i veličina paketa koje se šalju. Za igre koje se oslanjaju na igranjem mišem klijent bi slao puno više prometa prema poslužitelju, nego za igre koje se igraju isključivo pomoću tipkovnice. Korištenjem tehnika strojnog učenja iz mrežnih podataka moglo se s velikom preciznošću predvidjeti korisnička brzina kodiranja i broj okvira u sekundi, dok predviđanje rezolucije igre u najboljem slučaju je bilo točno u 70 % slučajeva.

Abstract (english)

Due to the ease of access, cloud gaming is gaining increasing popularity among gamers. In addition to that, it gives game publishers wider reach while also solving the problem of video game piracy. With the development of 5G networks, this form of gaming should be even more accessible.
This thesis analyzes the impact of bitrate, framerate and resolution on spatial and temporal video metrics, as well as their impact on network traffic. Machine learning was used to predict the resolution, bitrate and framerate from network data for each recorded session. When analyzing the influence of parameters on video metrics, it was determined that changing the resolution has a large impact on spatial metrics, while changing the framerate has an impact on temporal video metrics. Bitrate had no effect on either spatial or temporal metrics. It has also been observed that first and third person games have high temporal and low spatial metrics, while top down games have low temporal and high spatial video metric values. By analyzing network traffic, it was found that by increasing bitrate, the number of packets as well as packet size increased. For games that relied on mouse use, the client would send a lot more traffic data to the server than for games which were played solely using the keyboard. By using machine learning on traffic data, it is possible to predict bitrate and framerate with great accuracy, while predicting resolution was at best 70 % accurate.

• Keywords: Igre zasnovane na računalom oblak; Steam Remote Play; Strojno učenje; Metrike prostorne i vremenske složenosti videozapisa
• Keywords (english): Cloud gaming; Steam Remote Play; Machine learning; Spatial and temporal video metrics
• Language: croatian
• URN:NBN: urn:nbn:hr:168:188136
• Study programme: Title: Computing; Study programme type: university; Study level: graduate; Academic / professional title: magistar/magistra inženjer/inženjerka računarstva (magistar/magistra inženjer/inženjerka računarstva)
• Type of resource: Text
• File origin: Born digital
• Access conditions: Closed access
• Created on: 2023-01-14 09:38:26