Adversarial Training for Tabular Credit Scoring: A Multi-Attack Robustness Evaluation in P2P Lending

Paper: arXiv 2609.09945

Authors: Gijs A. F. Niewzwaag, Marijn G. S. Veth, Manuele Massei, Marcos R. Machado

Abstract

Machine learning-based credit scoring is increasingly central to Peer-to-Peer (P2P) lending, yet its resilience to adversarial manipulation, where applicants strategically alter self-reported inputs to secure favourable decisions, remains poorly understood. Most adversarial-robustness evidence comes from image and text domains and evaluates a single attack against a matching defence, offering little guidance on how defences generalise across attack types in tabular credit data. We address this with a systematic train-test robustness benchmark on a large Lending Club subset, spanning three model families (logistic regression, a feed-forward neural network, and a transformer for tabular data) and four attacks confined to applicant-mutable features: Fast Gradient Sign Method (FGSM), Projected Gradient Descent (PGD), Salt-and-Pepper (S&P) noise, and DeepFool, plus a mixed-attack regime. Across a full grid evaluated with stratified cross-validation, adversarial training sharply improves robustness against the attack it is trained on and transfers well within the gradient-based family, but transfers weakly to non-gradient corruption, so single-attack defences overstate real-world resilience. Mixed training delivers the most balanced robustness across heterogeneous attacks while preserving clean-test performance, supporting multi-attack stress testing in credit-model governance.

Complexity vs Empirical Score

• Math Complexity: 6.0/10
• Empirical Rigor: 8.0/10
• Quadrant: Holy Grail — high math complexity, high empirical rigor

Why this score: This paper presents a robust empirical evaluation of adversarial training in a critical financial application. Its strength lies in systematically testing multiple attack types and model families, offering valuable insights into real-world model resilience. The methodology is sound and the findings are clearly presented.

Research Flowchart

  flowchart TD
    A[Research Goal: Evaluate ML Model Robustness to Adversarial Attacks in Tabular Credit Scoring] --> B(Methodology: Systematic Train-Test Robustness Benchmark);
    B --> C{Data/Inputs: Large Lending Club Dataset};
    C --> D[Computational Processes: Evaluate 3 Model Families, 4 Attack Types (+ Mixed Attack), Adversarial Training, Stratified Cross-Validation];
    D --> E{Outcomes: Single-attack defenses overstate robustness; Mixed-attack training delivers balanced robustness & preserves clean performance.};

Cited References (88)

1. Regulation (EU) 2024/1361 of the European Parliament and of the Council of 13 March 2024 Laying Down Harmonised Rules on Artificial Intelligence (Artificial Intelligence Act) (2024) — link
2. The Global Findex Database 2021: Financial Inclusion, Digital Payments, and Resilience in the Age of COVID-19 (2022) — link
3. Enhancing credit scoring with alternative data (2021)
4. Economic Well-Being of U.S. Households in 2023 (2024) — link
5. Evaluating credit risk and loan performance in online Peer-to-Peer (P2P) lending (2015)
6. An econometric analysis of the bank credit scoring problem (1989)
7. Intelligent Credit Scoring: Building and Implementing Better Credit Risk Scorecards (2017)
8. An analytical approach to credit risk assessment using machine learning models (2025) — link

0 of 88 cited works are in the corpus; resolved via arXiv source bundles, Semantic Scholar and OpenAlex. Corpus-wide: 2518 linked papers.