Institutional Repository of Vilnius University
Saugus „Kubernetes“ paslapčių saugojimas ir tvarkymas „Git“ saugyklose /
Abstract
dc:descriptionThis bachelor's thesis investigates the challenges of secure management of sensitive information in Kubernetes environments under the GitOps methodology. The work covers two interconnected research objects. The first part analyses the Sealed Secrets tool, identifies its key shortcomings - the asynchronous validation flaw and the absence of an automated rotation mechanism - and proposes technically grounded enhancements: synchronous validation via a webhook mechanism and a cron-expression-based rotation policy. The improvements were implemented in a forked repository and validated by unit tests. The second part presents a machine-learning anomaly detection system for Kubernetes audit log analysis. The system includes a synthetic data generator producing three anomaly types (RBAC escalation, secrets-read spike, slow reconnaissance), a feature engineering module (seven features computed over 60-second tumbling windows), a comparison of four model families (LightGBM, XGBoost, Logistic Regression, Random Forest), and evaluation using time-series-aware cross-validation. Results show clear differentiation between models: tree ensembles achieve high precision (up to 0.934), while Logistic Regression achieves the best AUC-PR (0.875), enabling evidence-based recommendations for practical deployment.
Degree
thesis:*- Grantor dc:publisher
- Institutional Repository of Vilnius University
- Year dc:date
- 2026
Author and committee
dc:creator, dc:contributor.*- Author dc:creator
-
- Vertelis, Aldas,
Rights
dc:rights- Statement dc:rights
-
- info:eu-repo/semantics/openAccess
- Language dc:language
- lit
Identifiers
dc:identifier.*- Repository record dc:identifier
- https://repository.vu.lt/VU:ELABAETD308339090&prefLang=en_US