Jove
Visualize
Contact Us
JoVE
x logofacebook logolinkedin logoyoutube logo
ABOUT JoVE
OverviewLeadershipBlogJoVE Help Center
AUTHORS
Publishing ProcessEditorial BoardScope & PoliciesPeer ReviewFAQSubmit
LIBRARIANS
TestimonialsSubscriptionsAccessResourcesLibrary Advisory BoardFAQ
RESEARCH
JoVE JournalMethods CollectionsJoVE Encyclopedia of ExperimentsArchive
EDUCATION
JoVE CoreJoVE BusinessJoVE Science EducationJoVE Lab ManualFaculty Resource CenterFaculty Site
Terms & Conditions of Use
Privacy Policy
Policies

Related Experiment Videos

C/R-PMGAN: A dual-path self-supervised and patch-masked generative adversarial network for transferable black-box

Zihan Peng1, Yang Xu1, Sicong Zhang1

  • 1School of Cyber Science and Technology, Guizhou Normal University, Guiyang, 550001, Guizhou, China; Guizhou Key Laboratory of NewGen Cyberspace Security, Guiyang, 550001, Guizhou, China.

Neural Networks : the Official Journal of the International Neural Network Society
|March 7, 2026
PubMed
Summary

Related Concept Videos

Detection of Black Holes01:10

Detection of Black Holes

1.7K
Although black holes were theoretically postulated in the 1920s, they remained outside the domain of observational astronomy until the 1970s.
Their closest cousins are neutron stars, which are composed almost entirely of neutrons packed against each other, making them extremely dense. A neutron star has the same mass as the Sun but its diameter is only a few kilometers. Therefore, the escape velocity from their surface is close to the speed of light.
Not until the 1960s, when the first neutron...
1.7K
Masking and Demasking Agents01:19

Masking and Demasking Agents

4.1K
EDTA titrations may necessitate masking and demasking agents to temporarily protect a particular metal ion in a mixture from the EDTA reaction. These agents facilitate the sequential analysis of the metal ions by forming stable complexes with some—but not all—metal ions during certain steps.
There are many masking agents, such as cyanide, fluoride, triethanolamine, thiourea, and 2,3-bis(sulfanyl)propan-1-ol (formerly 2,3-dimercapto-1-propanol), with the masking agent chosen based on...
4.1K

You might also read

Related Articles

Articles linked to this work by shared authors, journal, and citation graph.

Sort by
Same author

Lower limb motion intention recognition using multi-source able-bodied gait signals.

Journal of neural engineering·2026
Same author

Domain-aware domain-class adaptation network for motor execution to motor imagery EEG classification.

Frontiers in neuroscience·2026
Same author

Noise-Driven Feature Enhancement of High-Frequency SSVEP Through Underdamped Second-Order Stochastic Resonance Energy Transfer.

IEEE transactions on neural systems and rehabilitation engineering : a publication of the IEEE Engineering in Medicine and Biology Society·2026
Same author

Molecular mechanisms of the MLL4 complex in H3K4 methylation and p53-dependent transcription activation.

Molecular cell·2026
Same author

Nonlinear Characteristics of High-Density Electroencephalography Demonstrate Differences in Neural Processing of Musical Syntax by Musicians and Nonmusicians.

Annals of the New York Academy of Sciences·2026
Same author

CAR T-cell therapy for patients with relapsed or refractory marginal zone lymphoma.

Lancet (London, England)·2026

This study introduces C/R-PMGAN, a novel algorithm enhancing adversarial attacks. It improves the efficiency and transferability of attacks across diverse models, crucial for real-world applications.

Area of Science:

  • Artificial Intelligence
  • Computer Vision
  • Machine Learning

Background:

  • Generative Adversarial Networks (GANs) excel at data generation and are used in black-box adversarial attacks.
  • Transferable black-box attacks are vital for real-world scenarios but face efficiency and overfitting issues with traditional methods.
  • Existing GAN-based methods struggle with cross-model feature alignment.

Purpose of the Study:

  • To propose C/R-PMGAN, an algorithm that enhances adversarial example transferability and improves efficiency.
  • To address limitations of traditional gradient-based and existing GAN-based transferable attacks.

Main Methods:

  • Introduced a dual-path self-supervised task (contrastive learning and rotation classification) in the discriminator for robust feature representation.
Keywords:
Generative adversarial network (GAN)Gradient aggregationRandom patch-maskingSelf-supervised learningTransferable black-box attack

Related Experiment Videos

  • Developed a gradient aggregation strategy using random patch-masking to focus perturbations on shared semantic information.
  • Focused on enhancing cross-model feature alignment and perturbation efficiency.

    • Main Results:

    • C/R-PMGAN achieved the highest average attack success rate across 16 diverse black-box models, including vision transformers and defense-trained models.
    • Outperformed state-of-the-art transferable attacks in generation efficiency.
    • Demonstrated superior perturbation imperceptibility compared to existing methods.

    Conclusions:

    • C/R-PMGAN effectively enhances the transferability and efficiency of adversarial attacks in black-box settings.
    • The proposed methods lead to more robust feature representations and targeted perturbations.
    • The algorithm shows significant promise for real-world adversarial attack applications.