Dependency-based anomaly detection and applications /
Files
(Published version)
Date
2021
Authors
Lu, Sha
Editors
Advisors
Journal Title
Journal ISSN
Volume Title
Type:
thesis
Citation
Statement of Responsibility
Conference Name
Abstract
Anomalies are patterns in data that do not conform to a well-defined notion of normal behavior and they often contain insightful information about the unusual behaviors or characteristics of the data generation process. This research has developed several novel anomaly detection techniques based on the dependency among variables to improve the efficiency, effectiveness, flexibility and interpretability of anomaly detection. Experiments with real-world data have shown the superiority of the proposed methods. To use the research to solve challenging real-world problems, we also apply the proposed approach to effectively detect adversarial attacks against deep neural networks on image data. In summary, this thesis has greatly advanced the research on anomaly detection by exploiting variable dependency from data.
School/Discipline
University of South Australia. UniSA STEM.
UniSA STEM
UniSA STEM
Dissertation Note
Thesis (PhD(Communication and Information Science))--University of South Australia, 2021.
Provenance
Copyright 2021 Sha Lu.
Description
1 ethesis (ix, 164 pages) :
illustrations (some colour)
Includes bibliographical references (pages 149-164)
illustrations (some colour)
Includes bibliographical references (pages 149-164)
Access Status
506 0#$fstar $2Unrestricted online access