Electronic Thesis/Dissertation
 

Leveraging Interpretable Machine Learning Approaches to Detect Anomaly Within UK Cloud Service Sector

Open Access Deposited

The increasing reliance on cloud-based platforms has saved time, money, and resources but also has lured criminals in cyber-attacks. With financial losses averaging £3.4 million per data breach, the need to identify cyber-attacks is critical. Anomaly detection models are designed to identify unusual activities including data breaches. However, anomaly detection models suffer from over-sensitivity despite their proved successes in various domains. This praxis addresses the problem by enhancing the trustworthiness of anomaly detection models. This praxis study employs a validation model that integrates Explainable Artificial Intelligence (XAI) with machine learning models. The methodology acquires and processes a dataset, trains supervised, semi-supervised and unsupervised anomaly detection models, and applies XAI technologies including SHapley Additive exPlanations (SHAP) and Local Interpretable Model-Agnostic Explanations (LIME) to generate interpretable results. The main findings reveal that XAI models can effectively validate machine learning predictions. XAI global explanations can validate whether dataset features are employed properly. Furthermore, XAI local explanations can validate whether the prediction result (true positive, false negative and etc.) is properly formed by the right dataset features. Utilizing these attributes, XAI models effectively explain the machine learning model predictions, provide transparency, show correlation between features and anomalous activities, and increase trust among stakeholders.

Author Language Keyword Date created Type of Work License
  • All rights reserved
Rights statement GW Unit Degree Advisor Committee Member(s) Persistent URL

Notice to Authors

If you are the author of this work and you have any questions about the information on this page, please use the Contact form to get in touch with us.

Thumbnail Title Date Uploaded Visibility Actions
Preview of Tong_gwu_0075A_16992.pdf Tong_gwu_0075A_16992.pdf 2025-04-09 Open Access