Back
45· Steady
Health & Science3h 59m ago
A SHAP-guided heterogeneous ensemble defense framework integrating heterogeneous architectures, PGD adversarial training, and SHAP-based routing was evaluated for financial risk assessment under white-box adversarial attacks.
Archive Window: 7 Days Left
Germany
Who
MLP, ResNet-1D, TabTransformer
What
A SHAP-guided heterogeneous ensemble defense framework integrating heterogeneous architectures, PGD adversarial training, and SHAP-based routing was evaluated for financial risk assessment under white-box adversarial attacks.
When
Sun, 04 Oct 2026 21:07:22 GMT · 3h 59m ago
Where
Germany ·
Why
Deep learning models in financial risk assessment are vulnerable to adversarial perturbations with economic and regulatory consequences.
The Frontline Impact
How this affects you
The framework achieved defended AUCs of 0.758 and 0.723 on German Credit and Lending Club datasets, respectively, and reduced the default-class attack success rate from an average of 0.535 to 0.20. SHAP analysis showed improved attribution consistency, increasing Spearman correlation between clean and defended explanations from 0.42 to 0.87.
Story chain
1 event in this threadNo related history yet - this is the origin event.