SAFE-OD: Drift-Aware Champion-Challenger Online Fraud Detection for Cloud-Native Payment Platforms
Global Media and Social Sciences Research Journal · 29.07.2026 · 7 (2) · 303-315
DOI: 10.71465/gmssrj212 ↗Kurz gefasst
Fachartikel in der Zeitschrift Global Media and Social Sciences Research Journal, online erschienen am 29.07.2026. Verfasst von Fabio Romano und 2 weiteren Autor:innen. Der Volltext ist frei zugänglich (Open Access, CC BY 4.0).
Abstract (Originalsprache)
Online payment fraud models operate under two conflicting requirements: they must adapt to changing transaction behavior, yet an unsafe update can increase false alerts or silently damage ranking quality. This paper presents SAFE-OD, a cloud-native champion-challenger design that separates synchronous scoring from asynchronous adaptation. A stable averaged stochastic-gradient logistic model remains on the request path; a Population Stability Index monitor raises drift suspicion; a plastic challenger is retrained from a rolling replay window; and a one-block delayed-label safety gate promotes the challenger only when its average precision exceeds the champion by 0.01 on the preceding block. We evaluate the design on the public ULB/Worldline dataset using a strict chronological split and a fixed top-0.5% alert budget. On the unmodified stream, SAFE-OD preserves the champion's average precision of 0.8209 and catches 185 of 223 frauds, whereas an unguarded challenger falls to 0.7116. Under a deterministic feature-representation drift stress test, SAFE-OD reaches 0.6558 average precision versus 0.5006 for the frozen champion and 0.4835 for naive continuous updating. The slower active linear scorer exhibits 73.71 microsecond p99 single-row latency on the evaluation host, while replay updates complete asynchronously in at most 69.85 ms. The results show that adaptation policy, not merely online model capability, is central to reliable fraud detection under drift.
Zitieren
APA
Romano, F., Nguyen, A., & Farouk, O. (2026). SAFE-OD: Drift-Aware Champion-Challenger Online Fraud Detection for Cloud-Native Payment Platforms. Global Media and Social Sciences Research Journal, 7(2), 303-315. https://doi.org/10.71465/gmssrj212
Vancouver
Romano F, Nguyen A, Farouk O. SAFE-OD: Drift-Aware Champion-Challenger Online Fraud Detection for Cloud-Native Payment Platforms. Global Media and Social Sciences Research Journal. 2026;7(2):303-315. doi:10.71465/gmssrj212
BibTeX
@article{Romano2026,
author = {Romano, Fabio and Nguyen, Alice and Farouk, Omar},
title = {SAFE-OD: Drift-Aware Champion-Challenger Online Fraud Detection for Cloud-Native Payment Platforms},
journal = {Global Media and Social Sciences Research Journal},
year = {2026},
volume = {7},
number = {2},
pages = {303--315},
doi = {10.71465/gmssrj212}
}
EUVIMED