Improving Fast Minimum-Norm Attacks with Hyperparameter Optimization

Giuseppe Floris, Raffaele Mura, Luca Scionis, Giorgio Piras, Maura Pintor +2 more
10/12/2023
cs.LGcs.CV

Official code: pralab/HOFMN (10 stars, MIT, last commit 2025-11-25).

Abstract

Evaluating the adversarial robustness of machine learning models using gradient-based attacks is challenging. In this work, we show that hyperparameter optimization can improve fast minimum-norm attacks by automating the selection of the loss function, the optimizer and the step-size scheduler, along with the corresponding hyperparameters. Our extensive evaluation involving several robust models demonstrates the improved efficacy of fast minimum-norm attacks when hyper-up with hyperparameter optimization. We release our open-source code at https://github.com/pralab/HO-FMN.

View on arXivView PDF

Code Implementations(1)

pralab/HOFMNOfficial100%
MIT
101Python, Jupyter NotebookMay 5, 202310 months agoMIT

Similar Papers

HO-FMN: Hyperparameter Optimization for Fast Minimum-Norm Attacks

95% match1 repo102024

Hyperparameter Learning under Data Poisoning: Analysis of the Influence of Regularization via Multiobjective Bilevel Optimization

89% match2023

Regularization properties of adversarially-trained linear regression

83% match1 repo42026

σ-zero: Gradient-based Optimization of $\ell_0$-norm Adversarial Examples

83% match2 repos1672024

Fast Adversarial Training against Sparse Attacks Requires Loss Smoothing

83% match2025

$L_p$-norm Distortion-Efficient Adversarial Attack

83% match2024

Cite this paper

@article{floris2023improving,
  title  = {Improving Fast Minimum-Norm Attacks with Hyperparameter Optimization},
  author = {Giuseppe Floris and Raffaele Mura and Luca Scionis and Giorgio Piras and Maura Pintor and Ambra Demontis and Battista Biggio},
  year   = {2023},
  eprint = {2310.08177},
  archivePrefix = {arXiv},
  url    = {http://arxiv.org/abs/2310.08177v1}
}

Add a badge to your README

Shows whether official code is linked for this paper.

Code of Paper badge
[![Code of Paper](https://codeofpaper.com/badge/2310.08177.svg)](https://codeofpaper.com/paper/2310.08177)

Discussion