Game-Theoretic Defenses for Robust Conformal Prediction Against Adversarial Attacks in Medical Imaging

2024-11-07

arXiv.org

AI Summary

This paper addresses the reliability and security of deep learning models for medical imaging under both known and unknown adversarial attacks. Methodologically, it proposes a robust uncertainty quantification framework integrating conformal prediction with zero-sum game theory: (1) an efficient conformal prediction set is constructed to ensure validity under known attacks; (2) a conservative threshold mechanism is designed to guarantee statistical coverage under unknown attacks; and (3) an optimal single-model defense strategy is derived within a zero-sum game formulation, ensuring theoretical convergence. The key contribution is the first integration of game-theoretic defense into conformal prediction, moving beyond static or uniform defense paradigms. Evaluated on three MedMNIST sub-datasets—including PathMNIST—the framework achieves nominal coverage exceeding 90%, significantly reduces prediction set size, and demonstrates stable superiority of the optimal game-theoretic strategy over baseline methods.

Technology Category

• Machine Learning: Adversarial Learning & Robustness
• Computer Vision: Adversarial Attacks & Robustness
• Game Theory and Economic Paradigms: Adversarial Learning

Application Category

• User Modeling, Personalization and Recommendation: Attacks and countermeasures in recommendation systems
• Security and Privacy: Large-scale security measurements
• Search and Retrieval-Augmented AI: Web learning to rank, online learning, and counterfactual learning for ranking

Abstract

Adversarial attacks pose significant threats to the reliability and safety of deep learning models, especially in critical domains such as medical imaging. This paper introduces a novel framework that integrates conformal prediction with game-theoretic defensive strategies to enhance model robustness against both known and unknown adversarial perturbations. We address three primary research questions: constructing valid and efficient conformal prediction sets under known attacks (RQ1), ensuring coverage under unknown attacks through conservative thresholding (RQ2), and determining optimal defensive strategies within a zero-sum game framework (RQ3). Our methodology involves training specialized defensive models against specific attack types and employing maximum and minimum classifiers to aggregate defenses effectively. Extensive experiments conducted on the MedMNIST datasets, including PathMNIST, OrganAMNIST, and TissueMNIST, demonstrate that our approach maintains high coverage guarantees while minimizing prediction set sizes. The game-theoretic analysis reveals that the optimal defensive strategy often converges to a singular robust model, outperforming uniform and simple strategies across all evaluated datasets. This work advances the state-of-the-art in uncertainty quantification and adversarial robustness, providing a reliable mechanism for deploying deep learning models in adversarial environments.

Problem

• Enhance model robustness against adversarial attacks in medical imaging.
• Construct valid conformal prediction sets under known and unknown attacks.
• Determine optimal defensive strategies using game-theoretic frameworks.

Innovation

• Integrates conformal prediction with game-theoretic defenses
• Uses maximum and minimum classifiers for defense aggregation
• Ensures coverage under unknown attacks via conservative thresholding