Calibrated Adversarial Sampling: Multi-Armed Bandit-Guided Generalization Against Unforeseen Attacks

Fuente: arXiv
Saved in:
Bibliographic Details
Main Authors: Wang, Rui, Wei, Zeming, Zhang, Xiyue, Sun, Meng
Format: Preprint
Published: 2025
Subjects:
Online Access:
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866914159991455744
author Wang, Rui
Wei, Zeming
Zhang, Xiyue
Sun, Meng
author_facet Wang, Rui
Wei, Zeming
Zhang, Xiyue
Sun, Meng
contents Deep Neural Networks (DNNs) are known to be vulnerable to various adversarial perturbations. To address the safety concerns arising from these vulnerabilities, adversarial training (AT) has emerged as one of the most effective paradigms for enhancing the robustness of DNNs. However, existing AT frameworks primarily focus on a single or a limited set of attack types, leaving DNNs still exposed to attack types that may be encountered in practice but not addressed during training. In this paper, we propose an efficient fine-tuning method called Calibrated Adversarial Sampling (CAS) to address these issues. From the optimization perspective within the multi-armed bandit framework, it dynamically designs rewards and balances exploration and exploitation by considering the dynamic and interdependent characteristics of multiple robustness dimensions. Experiments on benchmark datasets show that CAS achieves superior overall robustness while maintaining high clean accuracy, providing a new paradigm for robust generalization of DNNs.
format Preprint
id arxiv_https___arxiv_org_abs_2511_12265
institution arXiv
publishDate 2025
record_format arxiv
spellingShingle Calibrated Adversarial Sampling: Multi-Armed Bandit-Guided Generalization Against Unforeseen Attacks
Wang, Rui
Wei, Zeming
Zhang, Xiyue
Sun, Meng
Machine Learning
Artificial Intelligence
Cryptography and Security
Computer Vision and Pattern Recognition
Optimization and Control
Deep Neural Networks (DNNs) are known to be vulnerable to various adversarial perturbations. To address the safety concerns arising from these vulnerabilities, adversarial training (AT) has emerged as one of the most effective paradigms for enhancing the robustness of DNNs. However, existing AT frameworks primarily focus on a single or a limited set of attack types, leaving DNNs still exposed to attack types that may be encountered in practice but not addressed during training. In this paper, we propose an efficient fine-tuning method called Calibrated Adversarial Sampling (CAS) to address these issues. From the optimization perspective within the multi-armed bandit framework, it dynamically designs rewards and balances exploration and exploitation by considering the dynamic and interdependent characteristics of multiple robustness dimensions. Experiments on benchmark datasets show that CAS achieves superior overall robustness while maintaining high clean accuracy, providing a new paradigm for robust generalization of DNNs.
title Calibrated Adversarial Sampling: Multi-Armed Bandit-Guided Generalization Against Unforeseen Attacks
topic Machine Learning
Artificial Intelligence
Cryptography and Security
Computer Vision and Pattern Recognition
Optimization and Control
url https://arxiv.org/abs/2511.12265