Publicação
Improving Trustworthiness in DeFi: Anomaly Detection for Blockchain Oracle Price Feed Data
| datacite.subject.fos | Ciências Naturais::Ciências da Computação e da Informação | pt_PT |
| dc.contributor.advisor | Bação, Fernando José Ferreira Lucas | |
| dc.contributor.author | Gross, Lukas Sebastian | |
| dc.date.accessioned | 2024-11-11T15:18:48Z | |
| dc.date.available | 2024-11-11T15:18:48Z | |
| dc.date.issued | 2024-10-30 | |
| dc.description | Dissertation presented as the partial requirement for obtaining a Master's degree in Data Science and Advanced Analytics, specialization in Business Analytics | pt_PT |
| dc.description.abstract | This thesis investigates the application of anomaly detection techniques to blockchain oracle data, specifically focusing on the Ethereum - US Dollar exchange rate from the Chainlink oracle. The study addresses a critical gap in the reliability of decentralized finance applications by exploring both traditional static machine learning algorithms and novel streaming methods for identifying irregularities in oracle data feeds. The research compares the performance of various anomaly detection algorithms, including Support Vector Machines (SVM), DBSCAN, KMeans, Self-Organizing Maps (SOM), K-Nearest Neighbors (KNN), and Isolation Forest for static methods, and Sliding Window KNN (SWKNN), Robust Random Cut Forest (RRCF), and Half-Space Trees for streaming methods. These algorithms are evaluated using F1 score and recall as primary metrics, with a focus on their ability to detect sudden price spikes. Furthermore, the study examines the incorporating financial indicators such as Relative Strength Index, Average True Range, and Exponential Moving Average to enhance anomaly detection capabilities. The research utilizes UMAP projections for initial visual analysis and conducts comprehensive comparisons between the originally fetched univariate dataset and different multivariate datasets enhanced with additional features in form of financial indicators. The results indicate that multivariate datasets generally improve F1 scores, for static as well as streaming methods, while the streaming methods achieve a higher baseline on univariate datasets, compared to the static methods. This research contributes to the field by addressing the understudied aspect of anomaly detection in blockchain oracle data, providing insights into the effectiveness of various machine learning techniques in enhancing the security and reliability of DeFi applications. | pt_PT |
| dc.identifier.tid | 203784774 | pt_PT |
| dc.identifier.uri | http://hdl.handle.net/10362/174978 | |
| dc.language.iso | eng | pt_PT |
| dc.rights.uri | http://creativecommons.org/licenses/by/4.0/ | pt_PT |
| dc.subject | Blockchain | pt_PT |
| dc.subject | Smart Contracts | pt_PT |
| dc.subject | Decentralized Finance | pt_PT |
| dc.subject | Blockchain Oracles | pt_PT |
| dc.subject | Anomaly Detection | pt_PT |
| dc.subject | Streaming Data | pt_PT |
| dc.title | Improving Trustworthiness in DeFi: Anomaly Detection for Blockchain Oracle Price Feed Data | pt_PT |
| dc.type | master thesis | |
| dspace.entity.type | Publication | |
| rcaap.rights | openAccess | pt_PT |
| rcaap.type | masterThesis | pt_PT |
| thesis.degree.name | Mestrado em Ciência de Dados e Métodos Analíticos Avançados, especialização em Métodos Analíticos para a Gestão | pt_PT |