Full Text Available
Note: Clicking the button above will open the full text document at the original institutional repository in a new window.
Study of anomaly detection frameworks aiming for high performance
Jackson, L. A. 2025. Study of Anomaly Detection Frameworks Aiming for High Performance. Unpublished masters thesis. Stellenbosch: Stellenbosch University [online]. Available: https://scholar.sun.ac.za/items/403aeb9c-d37c-4f59-8bd0-5dffee47f17c
Saved in:
| Main Author: | |
|---|---|
| Other Authors: | |
| Format: | Thesis |
| Published: |
Stellenbosch : Stellenbosch University
2025
|
| Subjects: | |
| Tags: |
No Tags, Be the first to tag this record!
|
| _version_ | 1867613840540172288 |
|---|---|
| access_status_str | Open Access |
| author | Jackson, Laura Abigail |
| author2 | Von Leipzig, Konrad |
| author_browse | Jackson, Laura Abigail Von Leipzig, Konrad |
| author_facet | Von Leipzig, Konrad Jackson, Laura Abigail |
| author_sort | Jackson, Laura Abigail |
| collection | Thesis |
| dc_rights_str_mv | Stellenbosch University |
| description | Jackson, L. A. 2025. Study of Anomaly Detection Frameworks Aiming for High Performance. Unpublished masters thesis. Stellenbosch: Stellenbosch University [online]. Available: https://scholar.sun.ac.za/items/403aeb9c-d37c-4f59-8bd0-5dffee47f17c |
| format | Thesis |
| id | oai:scholar.sun.ac.za:10019.1/132485 |
| institution | Stellenbosch University (South Africa) |
| last_indexed | 2026-06-10T12:42:31.964Z |
| license_str | Other — see source repository |
| provenance_str_mv | Harvested via OAI-PMH from SUNScholar — Stellenbosch University Repository |
| publishDate | 2025 |
| publishDateRange | 2025 |
| publishDateSort | 2025 |
| publisher | Stellenbosch : Stellenbosch University |
| publisherStr | Stellenbosch : Stellenbosch University |
| record_format | dspace |
| source_str | SUNScholar — Stellenbosch University Repository |
| spelling | oai:scholar.sun.ac.za:10019.1/132485 Study of anomaly detection frameworks aiming for high performance Jackson, Laura Abigail Von Leipzig, Konrad Hummel, Vera Zincume, Philani Stellenbosch University. Faculty of Engineering. Dept. of Industrial Engineering. Anomaly detection (Computer security) Electronic data processing Data mining UCTD Jackson, L. A. 2025. Study of Anomaly Detection Frameworks Aiming for High Performance. Unpublished masters thesis. Stellenbosch: Stellenbosch University [online]. Available: https://scholar.sun.ac.za/items/403aeb9c-d37c-4f59-8bd0-5dffee47f17c Thesis (MEng)--Stellenbosch University, 2025. ENGLISH ABSTRACT: The research involves the development of an approach for high-performance anomaly detection, with consideration of existing anomaly detection frameworks. This research is driven by the need to address the lack of semi-automated anomaly detection frameworks that incorporate tools to address current challenges for anomaly detection in the industrial environment. The frameworks incorporate tools to address current anomaly detection challenges in the industrial environment, including the presence of noise, lack of labelled data, lack of universally applicable anomaly detection techniques, and the dynamic nature of normal and anomalous behaviour. Tool importance across different framework configurations is explored to understand the relative impact on anomaly detection capabilities. The key research question guiding this process is: “which combination of tools is necessary for high performance in anomaly detection to operate effectively and efficiently in the current industrial environment?”. The CRISP-DM methodology is used to guide the research, along with a study approach, including steps for the study development and results achieved during the study. The framework integrates state-of-the-art anomaly detection tools in an extensible, adaptable architecture designed to handle different diverse, large-scale datasets. The framework benefits include enhanced data quality, circumvention of challenges humans may encounter when attempting to identify anomalies, and saving time spent on the anomaly detection process, minimising downtime and disruptions. The framework is evaluated using an ablation study to understand the relative contribution of different tools to the overall framework anomaly detection capability with consideration of macro-average F1-score and execution time. This research contributes theoretical insights and practical applications of relevant tool capabilities and their combinations to guide the development of high-performance anomaly detection approaches. AFRIKAANSE OPSOMMING: Die navorsing behels die ontwikkeling van ‘n benadering vir ho¨e-prestasie-anomalie-opsporing, met inagneming van bestaaand anomalie-opsporingsraamwerke. Hierdie navorsing word gedryf deur die behoefde om die afwesigheid van semi-outomatiese anomalie opsorings raamwerke, wat tegnieke insluit om huidige uitdagings vir anomalie-opsporing in die industri¨ele omgewing te hanteer, aan te spreek. Die raamwerke sluit tegnieke in om huidige anomalie-opsporingsuitdagings in die industri¨ele omgewing aan te spreek, insluitende die teenwoordigheied van geraas, die gebrek aan gemerkde data, die gebrek aan universeel toepaslike anomalie-opsporings tegnieke, en die dinamiese aard van normale en abnormale gedrag. Die belangrikheid van tegnieke in verskillende raamwerkkonfigurasies word ondersoek om die relatiewe impak op anomalie-opsporingsvermo¨e te verstaan. Die kern-navorsingsvraag wat hierdie proses lei is: “watter kombinasie van relevante tegnieke, in ag genome data, word benodig om ‘n ho¨e-prestasie in anaomalie-opsporing te bewerkstellig om effektief en doeltreffend in die huidige industri¨ele omgewing te funksioneer?” Die “CRISP-DM” metodologie word gebruik om die navorsing te rig, tesame met ‘n gevallestudie benadering, insluitend stappe vir die studie-ontwikkeling en resultate wat tydens die studie bereik word, insluit. Die raamwerk integreer vooraanstaande anomalie-opsporingstegnieke in ‘n uitbreibare, aanpasbare argitektuur wat ontwerp is om verskillende diverse, grootskaalse datastelle te hanteer. Die voordele van die raamwerk sluit in verbeterde datakwalitiet, die oorbrugging van uitdagings wat mense mag te¨ekom wanneer hulle probeer om anomalie¨e te identifiseer, en tydsbesparing op die anomalie-opsporingsproses, wat stilstande en ontwrigtings minimaliseer. Die raamwerk word ge¨evalueer met behulp van ‘n ablasiestudie om die relatiewe bydrae van verskillende tegnieke tot die algehele raamwerk se anomalie-opsoringsvermo¨ens te verstaan, met inagneming van die makro-gemiddelde F1 telling en uitvoeringstyd. Hierdie navorsing dra by tot teoretiese insae en praktiese toepassings van relevante tegniekevermo¨ens en hul kombinasies om die ontwikkeling van ho¨e-prestasie anomalie-opsporingsbenaderings te rig. Masters 2025-06-09T13:43:04Z 2025-06-09T13:43:04Z 2025-03 Thesis https://scholar.sun.ac.za/handle/10019.1/132485 Stellenbosch University xxvi, 189 pages : illustrations application/pdf Stellenbosch : Stellenbosch University |
| spellingShingle | Anomaly detection (Computer security) Electronic data processing Data mining UCTD Jackson, Laura Abigail Study of anomaly detection frameworks aiming for high performance |
| title | Study of anomaly detection frameworks aiming for high performance |
| title_full | Study of anomaly detection frameworks aiming for high performance |
| title_fullStr | Study of anomaly detection frameworks aiming for high performance |
| title_full_unstemmed | Study of anomaly detection frameworks aiming for high performance |
| title_short | Study of anomaly detection frameworks aiming for high performance |
| title_sort | study of anomaly detection frameworks aiming for high performance |
| topic | Anomaly detection (Computer security) Electronic data processing Data mining UCTD |
| url | https://scholar.sun.ac.za/handle/10019.1/132485 |
| work_keys_str_mv | AT jacksonlauraabigail studyofanomalydetectionframeworksaimingforhighperformance |