Bridging Models to Defend: A Population-Based Strategy for Robust Adversarial Defense

Fuente: arXiv
Saved in:
Bibliographic Details
Main Authors: Wang, Ren, Li, Yuxuan, Chen, Can, Wang, Dakuo, Xiong, Jinjun, Chen, Pin-Yu, Liu, Sijia, Shahidehpour, Mohammad, Hero, Alfred
Format: Preprint
Published: 2023
Subjects:
Online Access:
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866909749865349120
author Wang, Ren
Li, Yuxuan
Chen, Can
Wang, Dakuo
Xiong, Jinjun
Chen, Pin-Yu
Liu, Sijia
Shahidehpour, Mohammad
Hero, Alfred
author_facet Wang, Ren
Li, Yuxuan
Chen, Can
Wang, Dakuo
Xiong, Jinjun
Chen, Pin-Yu
Liu, Sijia
Shahidehpour, Mohammad
Hero, Alfred
contents Adversarial robustness is a critical measure of a neural network's ability to withstand adversarial attacks at inference time. While robust training techniques have improved defenses against individual $\ell_p$-norm attacks (e.g., $\ell_2$ or $\ell_\infty$), models remain vulnerable to diversified $\ell_p$ perturbations. To address this challenge, we propose a novel Robust Mode Connectivity (RMC)-oriented adversarial defense framework comprising two population-based learning phases. In Phase I, RMC searches the parameter space between two pre-trained models to construct a continuous path containing models with high robustness against multiple $\ell_p$ attacks. To improve efficiency, we introduce a Self-Robust Mode Connectivity (SRMC) module that accelerates endpoint generation in RMC. Building on RMC, Phase II presents RMC-based optimization, where RMC modules are composed to further enhance diversified robustness. To increase Phase II efficiency, we propose Efficient Robust Mode Connectivity (ERMC), which leverages $\ell_1$- and $\ell_\infty$-adversarially trained models to achieve robustness across a broad range of $p$-norms. An ensemble strategy is employed to further boost ERMC's performance. Extensive experiments across diverse datasets and architectures demonstrate that our methods significantly improve robustness against $\ell_\infty$, $\ell_2$, $\ell_1$, and hybrid attacks. Code is available at https://github.com/wangren09/MCGR.
format Preprint
id arxiv_https___arxiv_org_abs_2303_10225
institution arXiv
publishDate 2023
record_format arxiv
spellingShingle Bridging Models to Defend: A Population-Based Strategy for Robust Adversarial Defense
Wang, Ren
Li, Yuxuan
Chen, Can
Wang, Dakuo
Xiong, Jinjun
Chen, Pin-Yu
Liu, Sijia
Shahidehpour, Mohammad
Hero, Alfred
Artificial Intelligence
Machine Learning
Neural and Evolutionary Computing
Adversarial robustness is a critical measure of a neural network's ability to withstand adversarial attacks at inference time. While robust training techniques have improved defenses against individual $\ell_p$-norm attacks (e.g., $\ell_2$ or $\ell_\infty$), models remain vulnerable to diversified $\ell_p$ perturbations. To address this challenge, we propose a novel Robust Mode Connectivity (RMC)-oriented adversarial defense framework comprising two population-based learning phases. In Phase I, RMC searches the parameter space between two pre-trained models to construct a continuous path containing models with high robustness against multiple $\ell_p$ attacks. To improve efficiency, we introduce a Self-Robust Mode Connectivity (SRMC) module that accelerates endpoint generation in RMC. Building on RMC, Phase II presents RMC-based optimization, where RMC modules are composed to further enhance diversified robustness. To increase Phase II efficiency, we propose Efficient Robust Mode Connectivity (ERMC), which leverages $\ell_1$- and $\ell_\infty$-adversarially trained models to achieve robustness across a broad range of $p$-norms. An ensemble strategy is employed to further boost ERMC's performance. Extensive experiments across diverse datasets and architectures demonstrate that our methods significantly improve robustness against $\ell_\infty$, $\ell_2$, $\ell_1$, and hybrid attacks. Code is available at https://github.com/wangren09/MCGR.
title Bridging Models to Defend: A Population-Based Strategy for Robust Adversarial Defense
topic Artificial Intelligence
Machine Learning
Neural and Evolutionary Computing
url https://arxiv.org/abs/2303.10225