Interpretable Identification of Dynamic Adaptive Streaming over HTTP (DASH) Flows Based on Feature Engineering

Internet service providers allocate network resources for different network flows. Among them, video streaming requires substantial network bandwidth to provide a satisfactory user experience. The identification of video traffic is one of the tools that helps to manage and optimise network resources...

Full description

Saved in:
Bibliographic Details
Main Author: Arkadiusz Biernacki
Format: Article
Language:English
Published: MDPI AG 2025-02-01
Series:Applied Sciences
Subjects:
Online Access:https://www.mdpi.com/2076-3417/15/5/2253
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1850050666373840896
author Arkadiusz Biernacki
author_facet Arkadiusz Biernacki
author_sort Arkadiusz Biernacki
collection DOAJ
description Internet service providers allocate network resources for different network flows. Among them, video streaming requires substantial network bandwidth to provide a satisfactory user experience. The identification of video traffic is one of the tools that helps to manage and optimise network resources. However, available solutions usually focus on traffic traces from a single application and use black-box models for identification, which require labels for training. To address this issue, we proposed an unsupervised machine learning model to identify traffic generated by video applications from the three popular services, namely YouTube, Netflix, and Amazon Prime. Our methodology involves feature generation, filtering, and clustering. The clustering used the most significant features to group similar traffic patterns. We employed the following three algorithms that represent different clustering methodologies: partition-based, density-based, and probabilistic approaches. The clustering achieved precision between 0.78 and 0.93, while recall rates ranged from 0.68 to 0.84, depending on the experiment parameters, which is comparable with black-box learning models. The model presented is interpretable and scalable, which is useful for its practical application.
format Article
id doaj-art-deda64d56d814b9dbb52f799d04e71c1
institution DOAJ
issn 2076-3417
language English
publishDate 2025-02-01
publisher MDPI AG
record_format Article
series Applied Sciences
spelling doaj-art-deda64d56d814b9dbb52f799d04e71c12025-08-20T02:53:22ZengMDPI AGApplied Sciences2076-34172025-02-01155225310.3390/app15052253Interpretable Identification of Dynamic Adaptive Streaming over HTTP (DASH) Flows Based on Feature EngineeringArkadiusz Biernacki0Department of Computer Networks and Systems, Silesian University of Technology, Akademicka 16, 44-100 Gliwice, PolandInternet service providers allocate network resources for different network flows. Among them, video streaming requires substantial network bandwidth to provide a satisfactory user experience. The identification of video traffic is one of the tools that helps to manage and optimise network resources. However, available solutions usually focus on traffic traces from a single application and use black-box models for identification, which require labels for training. To address this issue, we proposed an unsupervised machine learning model to identify traffic generated by video applications from the three popular services, namely YouTube, Netflix, and Amazon Prime. Our methodology involves feature generation, filtering, and clustering. The clustering used the most significant features to group similar traffic patterns. We employed the following three algorithms that represent different clustering methodologies: partition-based, density-based, and probabilistic approaches. The clustering achieved precision between 0.78 and 0.93, while recall rates ranged from 0.68 to 0.84, depending on the experiment parameters, which is comparable with black-box learning models. The model presented is interpretable and scalable, which is useful for its practical application.https://www.mdpi.com/2076-3417/15/5/2253video traffic identificationtraffic clusteringadaptive video
spellingShingle Arkadiusz Biernacki
Interpretable Identification of Dynamic Adaptive Streaming over HTTP (DASH) Flows Based on Feature Engineering
Applied Sciences
video traffic identification
traffic clustering
adaptive video
title Interpretable Identification of Dynamic Adaptive Streaming over HTTP (DASH) Flows Based on Feature Engineering
title_full Interpretable Identification of Dynamic Adaptive Streaming over HTTP (DASH) Flows Based on Feature Engineering
title_fullStr Interpretable Identification of Dynamic Adaptive Streaming over HTTP (DASH) Flows Based on Feature Engineering
title_full_unstemmed Interpretable Identification of Dynamic Adaptive Streaming over HTTP (DASH) Flows Based on Feature Engineering
title_short Interpretable Identification of Dynamic Adaptive Streaming over HTTP (DASH) Flows Based on Feature Engineering
title_sort interpretable identification of dynamic adaptive streaming over http dash flows based on feature engineering
topic video traffic identification
traffic clustering
adaptive video
url https://www.mdpi.com/2076-3417/15/5/2253
work_keys_str_mv AT arkadiuszbiernacki interpretableidentificationofdynamicadaptivestreamingoverhttpdashflowsbasedonfeatureengineering