Benchmark RadarFraudBench
A protocol-sensitive benchmark for adversarial robustness in financial fraud and credit risk, evaluating the same model-attack-defence setting under three protocols: unconstrained, post-hoc filtering, and deployment-aware constraint-integrated attacks.
- Released
- 2026-08-25
- Readiness
- Paper only
- Primary field
- Finance & Economics
Why it matters
Highlights that robustness conclusions are protocol-dependent, urging joint reporting of predictive degradation and attack feasibility, and incorporation of domain constraints into attack generation.
Motivation
Machine learning models are widely used in financial fraud and credit-risk detection, yet their adversarial robustness remains difficult to evaluate because financial tabular data involve domain-specific constraints, severe class imbalance, and asymmetric attacker capability.
Primary resources
Benchmark Radar records only publicly supported details and links back to primary sources for verification.