Research map / Adversarial robustness and security / Black-box and transfer attacks
Mechanistic interventions for explainable digital pathology uncovers adversarial vulnerabilities
Arijit Patra, Samiran Dey, Ajitha Rajan, Greg Slabaugh, Tapabrata Chakraborti
Read the paperCheck your idea against 137,217 papers
AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.
Measured effects reported in the paper
| What was changed | Where | Measured | Effect |
|---|---|---|---|
| add circuit-specific regularization (alpha=0.01) to adversarial training (ablation) | MIAT · TCGA-Prostate / BreastPathQ | adversarial accuracy: robustness dropped to the levels achieved by standard AT | Improved |
| target the patched vulnerable circuits instead of randomly selected circuits (ablation) | MIAT · TCGA-Prostate / BreastPathQ | adversarial accuracy: no improvement over standard AT | Improved |
| replace standard AT with MIAT | ViT-B/16 · TCGA-Prostate | PGD accuracy: 68.4 -> 75.3 | Improved |
| replace standard AT with MIAT | ViT-B/16 · TCGA-Prostate | CW accuracy: 55.2 -> 62.8 | Improved |
| replace standard AT with MIAT | ResNet-50 · TCGA-Prostate | PGD accuracy: 62.1 -> 70.5 | Improved |
| replace standard AT with MIAT | ResNet-50 · TCGA-Prostate | CW accuracy: 50.3 -> 58.9 | Improved |
| replace standard AT with MIAT | ViT-B/16 · TCGA-Prostate | clean accuracy: 85.1 -> 86.8 | Improved |
| apply adversarial training (AT) to clean-trained model | ViT-B/16 · TCGA-Prostate | clean accuracy: 88.2 -> 85.1 | Hurt |
| use MIAT instead of LAT | ViT-B/16 · TCGA-Prostate | PGD accuracy: 70.1 -> 75.3 | Improved |
| use MIAT instead of LAT | ViT-B/16 · TCGA-Prostate | CW accuracy: 57.3 -> 62.8 | Improved |
| use MIAT instead of TRADES | ViT-B/16 · TCGA-Prostate | PGD accuracy: 65.2 -> 75.3 | Improved |
| use MIAT instead of MART | ViT-B/16 · TCGA-Prostate | PGD accuracy: 67.5 -> 75.3 | Improved |
Closest papers by meaning
- Attacking CNNs in Histopathology with SNAP: Sporadic and Naturalistic Adversarial Patches (Student Abstract) (AAAI 2024 · similarity 0.87)
- Patho-R1: A Multimodal Reinforcement Learning-Based Pathology Expert Reasoner (AAAI 2026 · similarity 0.84)
- NaFV-Net: An Adversarial Four-view Network for Mammogram Classification (AAAI 2025 · similarity 0.84)
- PathAsst: A Generative Foundation AI Assistant towards Artificial General Intelligence of Pathology (AAAI 2024 · similarity 0.83)
- Prompt2Perturb (P2P): Text-Guided Diffusion-Based Adversarial Attack on Breast Ultrasound Images (CVPR 2025 · similarity 0.83)
- Symb-xMIL: Symbolic Explanations for Multiple Instance Learning in Digital Pathology (NeurIPS 2026 · similarity 0.83)
Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.
