We gratefully acknowledge support from
the Simons Foundation and member institutions.
Full-text links:

Download:

Current browse context:

cs.LG

Change to browse by:

References & Citations

DBLP - CS Bibliography

Bookmark

(what is this?)
CiteULike logo BibSonomy logo Mendeley logo del.icio.us logo Digg logo Reddit logo ScienceWISE logo

Computer Science > Machine Learning

Title: Fuzziness-tuned: Improving the Transferability of Adversarial Examples

Abstract: With the development of adversarial attacks, adversairal examples have been widely used to enhance the robustness of the training models on deep neural networks. Although considerable efforts of adversarial attacks on improving the transferability of adversarial examples have been developed, the attack success rate of the transfer-based attacks on the surrogate model is much higher than that on victim model under the low attack strength (e.g., the attack strength $\epsilon=8/255$). In this paper, we first systematically investigated this issue and found that the enormous difference of attack success rates between the surrogate model and victim model is caused by the existence of a special area (known as fuzzy domain in our paper), in which the adversarial examples in the area are classified wrongly by the surrogate model while correctly by the victim model. Then, to eliminate such enormous difference of attack success rates for improving the transferability of generated adversarial examples, a fuzziness-tuned method consisting of confidence scaling mechanism and temperature scaling mechanism is proposed to ensure the generated adversarial examples can effectively skip out of the fuzzy domain. The confidence scaling mechanism and the temperature scaling mechanism can collaboratively tune the fuzziness of the generated adversarial examples through adjusting the gradient descent weight of fuzziness and stabilizing the update direction, respectively. Specifically, the proposed fuzziness-tuned method can be effectively integrated with existing adversarial attacks to further improve the transferability of adverarial examples without changing the time complexity. Extensive experiments demonstrated that fuzziness-tuned method can effectively enhance the transferability of adversarial examples in the latest transfer-based attacks.
Subjects: Machine Learning (cs.LG); Cryptography and Security (cs.CR); Computer Vision and Pattern Recognition (cs.CV)
Cite as: arXiv:2303.10078 [cs.LG]
  (or arXiv:2303.10078v1 [cs.LG] for this version)

Submission history

From: Xiangyuan Yang [view email]
[v1] Fri, 17 Mar 2023 16:00:18 GMT (1990kb,D)

Link back to: arXiv, form interface, contact.