FinDEx: A Synthetic Data Sharing Platform for Financial Fraud Detection

dc.contributor.authorKarst, Fabian
dc.contributor.authorLi, Mahei
dc.contributor.authorLeimeister, Jan
dc.date.accessioned2023-12-26T18:43:32Z
dc.date.available2023-12-26T18:43:32Z
dc.date.issued2024-01-03
dc.identifier.doi10.24251/HICSS.2023.513
dc.identifier.isbn978-0-9981331-7-1
dc.identifier.other35ff90db-8c26-412c-ace0-e0cf663c196b
dc.identifier.urihttps://hdl.handle.net/10125/106897
dc.language.isoeng
dc.relation.ispartofProceedings of the 57th Hawaii International Conference on System Sciences
dc.rightsAttribution-NonCommercial-NoDerivatives 4.0 International
dc.rights.urihttps://creativecommons.org/licenses/by-nc-nd/4.0/
dc.subjectDesigning Data Ecosystems: Value, Impacts, and Fundamentals
dc.subjectdata ecosystem
dc.subjectdata scarcity
dc.subjectfinancial services
dc.subjectfraud detection
dc.subjectsynthetic data
dc.titleFinDEx: A Synthetic Data Sharing Platform for Financial Fraud Detection
dc.typeConference Paper
dc.type.dcmiText
dcterms.abstractThe rising number of financial frauds inflicted in the last year more than 800 billion USD in damages on the global economy. Although financial institutions possess advanced AI systems for fraud detection, the time required to accumulate a sufficient volume of fraudulent data for training models creates a costly vulnerability. Combined with the inability to share fraud detection training data among institutions due to data and privacy regulations, this poses a major challenge. To address this issue, we propose the concept of a synthetic data-sharing ecosystem platform (FinDEx). This platform ensures data anonymity by generating synthesized training data based on each institution's fraud detection datasets. Various synthetic data generation techniques are employed to rapidly construct a shared dataset for all ecosystem members. Using design science research, this paper leverages insights from financial fraud detection literature, data sharing practices, and modular systems theory to derive design knowledge for the platform architecture. Furthermore, the feasibility of using different data generation algorithms such as generative adversarial networks, variational auto encoder and Gaussian mixture model was evaluated and different methods for the integration of synthetic data into the training procedure were tested. Thus, contributing to the theory at the intersection between fraud detection and data sharing and providing practitioners with guidelines on how to design such systems.
dcterms.extent10 pages
prism.startingpage4258

Files

Original bundle
Now showing 1 - 1 of 1
Loading...
Thumbnail Image
Name:
0420.pdf
Size:
2.37 MB
Format:
Adobe Portable Document Format