Home // DATA ANALYTICS 2025, The Fourteenth International Conference on Data Analytics // View article

Achieving Near Real-Time Data Freshness in Fraud Detection: An HTAP Approach

Authors:
Joseph G. Vella
Matteo G. Giorgino

Keywords: HTAP System; Near Real-Time Fraud Detection; Database Architecture; Data Freshness; DBMS Benchmarking.

Abstract:
The rise of complex financial fraud in banking demands sophisticated detection solutions capable of near-real-time operations, delivering rapid responses on fresh data. Traditional architectures that connect Online Transactional Processing (OLTP) and Online Analytical Processing (OLAP) through Extract-Transform-Load (ETL) pipelines often fail to satisfy these requirements, particularly when both data consistency and rapid response times are critical. This paper examines how Hybrid Transactional/Analytical Processing (HTAP) architectures can address these limitations by consolidating transactional and analytical workloads within a single system. To assess HTAP’s suitability for Fraud Detection in near-real-time scenarios, the paper employs HyBench, a benchmarking framework that measures data freshness in centralised HTAP systems, augmented with a custom-made external harness. This setup allows for systematic scenario exploration and detailed performance tracking under realistic banking workloads. Across 48 hours, the evaluation executes 96 parameterised runs at multiple data volumes, with database configurations optimised for the available hardware. Results indicate that an HTAP platform can sustain continuous access to fresh data, achieving sub-20 ms freshness, even under mixed OLTP and OLAP loads, while maintaining high transactional throughput. Although there are efficiency trade-offs compared to standalone OLTP or OLAP deployments, proper system configuration and tuning prove critical for balancing performance and freshness. Furthermore, the flexible benchmarking harness developed here enables practitioners to define custom metrics and integrate additional processing logic into the pipeline, extending beyond HyBench’s capabilities.

Pages: 18 to 23

Copyright: Copyright (c) IARIA, 2025

Publication date: September 28, 2025

Published in: conference

ISSN: 2308-4464

ISBN: 978-1-68558-293-7

Location: Lisbon, Portugal

Dates: from September 28, 2025 to October 2, 2025