mlsanityData to 14 Oct 2026

Improving Integrated Gradient-based Transferable Adversarial Examples by Refining the Integration Path

arXiv 2412.18844 ↗First posted 25 Dec 2024cs.CR cs.CV cs.LGcited by 10

Not on the map. Why?

Abstract (arXiv)

Transferable adversarial examples are known to cause threats in practical, black-box attack scenarios. A notable approach to improving transferability is using integrated gradients (IG), originally developed for model interpretability. In this paper, we find that existing IG-based attacks have limited transferability due to their naive adoption of IG in model interpretability. To address this limitation, we focus on the IG integration path and refine it in three aspects: multiplicity, monotonicity, and diversity, supported by theoretical analyses. We propose the Multiple Monotonic Diversified Integrated Gradients (MuMoDIG) attack, which can generate highly transferable adversarial examples on different CNN and ViT models and defenses. Experiments validate that MuMoDIG outperforms the latest IG-based attack by up to 37.3% and other state-of-the-art attacks by 8.4%. In general, our study reveals that migrating established techniques to improve transferability may require non-trivial efforts. Code is available at urlhttps://github.com/RYC-98/MuMoDIG.

Key facts. Cited by 10 papers in the map (data through 14 Oct 2026). Builds on 12 of its 44 references that are in the map. First posted on arXiv 25 Dec 2024.

Builds on

12 of 44 references.

References · 12

  1. 2020An Image is Worth 16x16 Words: Transformers for Image Recognition at Scalecited by 30,186
  2. 2017MobileNets: Efficient Convolutional Neural Networks for Mobile Vision Applicationscited by 3,845
  3. 2017Axiomatic Attribution for Deep Networkscited by 3,384
  4. 2017Ensemble Adversarial Training: Attacks and Defensescited by 840
  5. 2016Delving into Transferable Adversarial Examples and Black-box Attackscited by 536
  6. 2017Countering Adversarial Images using Input Transformationscited by 520
  7. 2017Decision-Based Adversarial Attacks: Reliable Attacks Against Black-Box Machine Learning Modelscited by 462
  8. 2020Skip Connections Matter: On the Transferability of Adversarial Examples Generated with ResNetscited by 167
  9. 2020On Success and Simplicity: A Second Look at Transferable Targeted Attackscited by 89
  10. 2023Content-based Unrestricted Adversarial Attackcited by 73
  11. 2022Transferable Adversarial Attack based on Integrated Gradientscited by 35
  12. 2023Revisiting Transferable Adversarial Images: Systemization, Evaluation, and New Insightscited by 1

Built on by

Cited by year (2026 to 14 Oct)

Citing papers per year: 2025: 5; 2026: 5.

10 papers.

  1. 2025Improving Adversarial Transferability on Vision Transformers via Forward Propagation Refinementcited by 7
  2. 2025Boosting the Local Invariance for Better Adversarial Transferabilitycited by 2
  3. 2025Path-Weighted Integrated Gradients for Interpretable Dementia Classificationcited by 1
  4. 2026The Power of Decaying Steps: Enhancing Attack Stability and Transferability for Sign-based Optimizerscited by 1
  5. 2025Latent Danger Zone: Distilling Unified Attention for Cross-Architecture Black-box Attackscited by 0
  6. 2025Beyond Deceptive Flatness: Dual-Order Solution for Strengthening Adversarial Transferabilitycited by 0
  7. 2026Devling into Adversarial Transferability on Image Classification: Review, Benchmark, and Evaluationcited by 0
  8. 2026FRInGe: Distribution-Space Integrated Gradients with Fisher--Rao Geometrycited by 0
  9. 2026Towards Universal Physical Adversarial Attacks via a Joint Multi-Objective and Multi-Model Optimization Frameworkcited by 0
  10. 2026On Success and Simplicity: A Second Look at Transferable Vision-Language Attack Pipelinecited by 0