Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

MOS-Attack: A Scalable Multi-objective Adversarial Attack Framework

Ping Guo, Cheng Gong, Xi Lin, Fei Liu, Zhichao Lu, Qingfu Zhang, Zhenkun Wang

TL;DR

The paper tackles the problem of evaluating DNN robustness with adversarial examples by unifying multiple surrogate losses into a scalable, parameter-free framework. It introduces MOS Attack, a smooth set-based multi-objective optimization approach that coordinates several loss functions and automatically mines synergistic patterns to reduce the number of needed objectives. Empirically, MOS-8 achieves state-of-the-art-like performance on CIFAR-10 and ImageNet and outperforms single-objective baselines, while MOS-3$^*$ demonstrates that a compact, synergy-driven attack can match much of the effectiveness with fewer losses. The work provides both a practical, resource-efficient adversarial framework and valuable insights into loss interactions, with an available codebase for reproduction and extension.

Abstract

Crafting adversarial examples is crucial for evaluating and enhancing the robustness of Deep Neural Networks (DNNs), presenting a challenge equivalent to maximizing a non-differentiable 0-1 loss function. However, existing single objective methods, namely adversarial attacks focus on a surrogate loss function, do not fully harness the benefits of engaging multiple loss functions, as a result of insufficient understanding of their synergistic and conflicting nature. To overcome these limitations, we propose the Multi-Objective Set-based Attack (MOS Attack), a novel adversarial attack framework leveraging multiple loss functions and automatically uncovering their interrelations. The MOS Attack adopts a set-based multi-objective optimization strategy, enabling the incorporation of numerous loss functions without additional parameters. It also automatically mines synergistic patterns among various losses, facilitating the generation of potent adversarial attacks with fewer objectives. Extensive experiments have shown that our MOS Attack outperforms single-objective attacks. Furthermore, by harnessing the identified synergistic patterns, MOS Attack continues to show superior results with a reduced number of loss functions. Our code is available at https://github.com/pgg3/MOS-Attack.

MOS-Attack: A Scalable Multi-objective Adversarial Attack Framework

TL;DR

The paper tackles the problem of evaluating DNN robustness with adversarial examples by unifying multiple surrogate losses into a scalable, parameter-free framework. It introduces MOS Attack, a smooth set-based multi-objective optimization approach that coordinates several loss functions and automatically mines synergistic patterns to reduce the number of needed objectives. Empirically, MOS-8 achieves state-of-the-art-like performance on CIFAR-10 and ImageNet and outperforms single-objective baselines, while MOS-3 demonstrates that a compact, synergy-driven attack can match much of the effectiveness with fewer losses. The work provides both a practical, resource-efficient adversarial framework and valuable insights into loss interactions, with an available codebase for reproduction and extension.

Abstract

Crafting adversarial examples is crucial for evaluating and enhancing the robustness of Deep Neural Networks (DNNs), presenting a challenge equivalent to maximizing a non-differentiable 0-1 loss function. However, existing single objective methods, namely adversarial attacks focus on a surrogate loss function, do not fully harness the benefits of engaging multiple loss functions, as a result of insufficient understanding of their synergistic and conflicting nature. To overcome these limitations, we propose the Multi-Objective Set-based Attack (MOS Attack), a novel adversarial attack framework leveraging multiple loss functions and automatically uncovering their interrelations. The MOS Attack adopts a set-based multi-objective optimization strategy, enabling the incorporation of numerous loss functions without additional parameters. It also automatically mines synergistic patterns among various losses, facilitating the generation of potent adversarial attacks with fewer objectives. Extensive experiments have shown that our MOS Attack outperforms single-objective attacks. Furthermore, by harnessing the identified synergistic patterns, MOS Attack continues to show superior results with a reduced number of loss functions. Our code is available at https://github.com/pgg3/MOS-Attack.
Paper Structure (19 sections, 9 equations, 4 figures, 6 tables, 1 algorithm)

This paper contains 19 sections, 9 equations, 4 figures, 6 tables, 1 algorithm.

Table of Contents

  1. Introduction
  2. Background
  3. Multi-Objective Set-based Attack
  4. Multi-Objective Adversarial Attack
  5. Decomposition-Based Optimization
  6. Smooth Set-based Optimization
  7. Methodology
  8. MOS Attack: Implementation by APGD
  9. Automated Synergistic Pattern Mining
  10. Implementation: Loss Functions
  11. Runtime/Efficiency Comparison
  12. Experiment
  13. Experiment Setup
  14. Overall Results
  15. MOS Attack Upper Bound
  16. ...and 4 more sections

Figures (4)

  • Figure 1: Comparison of different optimization methods for conducting multi-objective adversarial attacks.
  • Figure 1: The loss function utilized for implementing our attack.
  • Figure 2: Occurrences of different loss synergistic patterns across CIFAR-10 and ImageNet datasets. We only retain the top patterns that account for more than $1\%$ of the adversarial examples.
  • Figure 3: Detailed distribution of loss synergistic patterns across different model architectures. We only retain the top patterns that account for more than $1\%$ of the adversarial examples.

Theorems & Definitions (2)

  • Definition 3.1: Pareto Optimal
  • Definition 3.2: Pareto Set and Pareto Front