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

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.

On the tradeoff between robustness and geometric accuracy in adversarial training
Rachel Morris, Ryan Murray, Kerrek Stinson
arXiv16 Oct 2026PreprintCC BY 4.0Full textAnalysis & PDEsarXiv
Assessing Adversarial Robustness of Latent Reasoning Models
Shaolong Chen, Ang Li, Mingjie Li +1
arXiv3 Sep 2026PreprintCC BY 4.0Full textStatisticsarXiv
Sequential Euclidean connections with exponential memory: distributional performance and adversarial robustness
Pedro M. M. de Castro
arXiv29 Aug 2026PreprintarXiv licenceFull textOptimisation & controlarXiv
Statistical Inference for Adversarial Training: Central Limit Theorems via Optimal Transport
Kim Jakwang, Kwon Dohyun
arXiv4 Sep 2026PreprintCC BY 4.0Full textProbability StatisticsarXiv
Below what training size do deep tabular generators stop beating trivial baselines? A preregistered benchmark on a size ladder of clinical and standard datasets
Shivam Shrivastava
arXiv2 Oct 2026PreprintCC BY 4.0Full textStatisticsarXiv

Cite

BibTeX RIS

BibTeX

@misc{niewzwaag2026adversarial,
  title = {{Adversarial Training for Tabular Credit Scoring: A Multi-Attack Robustness Evaluation in P2P Lending}},
  author = {Gijs A. F. Niewzwaag and Marijn G. S. Veth and Manuele Massei and Marcos R. Machado},
  howpublished = {arXiv},
  year = {2026},
  month = sep,
  doi = {10.48550/arxiv.2609.09945},
  eprint = {2609.09945},
  archivePrefix = {arXiv},
  url = {https://arxiv.org/abs/2609.09945},
  note = {CC BY 4.0},
}

RIS

TY  - UNPB
TI  - Adversarial Training for Tabular Credit Scoring: A Multi-Attack Robustness Evaluation in P2P Lending
AU  - Gijs A. F. Niewzwaag
AU  - Marijn G. S. Veth
AU  - Manuele Massei
AU  - Marcos R. Machado
PY  - 2026
DA  - 2026/09/09
PB  - arXiv
DO  - 10.48550/arxiv.2609.09945
UR  - https://arxiv.org/abs/2609.09945
AB  - 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.
ER  -

Abstract

Related

Cite