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

Attentional semantic attack for enhancing adversarial samples transferability.

Pengju Wang1, Jing Liu2

  • 1College of Computer Science, Inner Mongolia University, Hohhot, 010010, China.

Scientific Reports
|March 27, 2026
PubMed
Summary
This summary is machine-generated.

Related Concept Videos

Improving Translational Accuracy02:07

Improving Translational Accuracy

3.8K
3.8K
Improving Translational Accuracy02:07

Improving Translational Accuracy

15.4K
Base complementarity between the three base pairs of mRNA codon and the tRNA anticodon is not a failsafe mechanism. Inaccuracies can range from a single mismatch to no correct base pairing at all. The free energy difference between the correct and nearly correct base pairs can be as small as 3 kcal/ mol. With complementarity being the only proofreading step, the estimated error frequency would be one wrong amino acid in every 100 amino acids incorporated. However, error frequencies observed in...
15.4K

You might also read

Related Articles

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

Sort by
Same author

[Isolation of functional bacteria guided by PCR-DGGE technology from high temperature petroleum reservoirs].

Huan jing ke xue= Huanjing kexue·2008
Same author

Assessing the spatial extent of breast tumor intrinsic optical contrast using ultrasound and diffuse optical spectroscopy.

Journal of biomedical optics·2008
Same author

alpha4/7-conotoxin Lp1.1 is a novel antagonist of neuronal nicotinic acetylcholine receptors.

Peptides·2008
Same author

[Construction and characterization of avian pathogenic Escherichia coli mutants with iro and/or tsh gene mutation].

Sheng wu gong cheng xue bao = Chinese journal of biotechnology·2008
Same author

[Enhancement of GFP expression by Kozak sequence +4G in HEK293 cells].

Sheng wu gong cheng xue bao = Chinese journal of biotechnology·2008
Same author

Sphingomyelin synthase 2 deficiency attenuates NFkappaB activation.

Arteriosclerosis, thrombosis, and vascular biology·2008
Same journal

Correction: A method for supervoxel-wise association studies of age and other non-imaging variables from coronary computed tomography angiograms.

Scientific reports·2026
Same journal

Poly(bromophenol blue)/CoSn(OH)<sub>6</sub> cubic particles modified pencil graphite electrode for electrochemical determination of diphenhydramine.

Scientific reports·2026
Same journal

Dietary Chlorella, Spirulina, and acidifier modulate jejunal cytokine-related gene expression in broiler chickens.

Scientific reports·2026
Same journal

Perceived physical activity barriers in university students: associations with fatigue and eating behaviours.

Scientific reports·2026
Same journal

Refuge limitation structures habitat use in agricultural landscapes: evidence from Sunda pangolins.

Scientific reports·2026
Same journal

Lightweight stateless transaction verification with outsourced witness updates for UTXO blockchains.

Scientific reports·2026
See all related articles

Deep neural networks (DNNs) are vulnerable to adversarial attacks. This study introduces Attentional Semantic Attack (ASA), a novel method that improves adversarial attack transferability across different models.

Area of Science:

  • Artificial Intelligence
  • Machine Learning
  • Computer Vision

Background:

  • Deep neural networks (DNNs) are susceptible to adversarial attacks, leading to incorrect outputs.
  • Existing black-box attacks have limited transferability due to model differences.
  • Attention mechanisms in DNNs exhibit similar semantic properties across models for identical inputs.

Purpose of the Study:

  • To propose a novel adversarial attack method that enhances transferability in black-box settings.
  • To leverage the shared attentional semantic properties of different DNNs for more effective attacks.
  • To introduce Attentional Semantic Attack (ASA) for inducing misclassifications across multiple models.

Main Methods:

  • Developed "attentional perturbation" to capture and modify shared attentional semantic properties.
Keywords:
Adversarial attackAttentional perturbationTransferability

Related Experiment Videos

  • Proposed the Attentional Semantic Attack (ASA) method utilizing these perturbations.
  • Optimized a loss function to degrade the attentional semantic property of samples, inducing misclassification.
  • Iteratively added perturbations to generate adversarial samples with high transferability.

    • Main Results:

    • ASA demonstrates superior transferability compared to state-of-the-art adversarial attacks.
    • Experiments on the ImageNet dataset validate the effectiveness of ASA.
    • The method successfully induces misclassifications by degrading shared attentional semantic properties.

    Conclusions:

    • Attentional perturbation is an effective strategy for enhancing adversarial attack transferability.
    • ASA offers a promising approach for robust black-box adversarial attacks.
    • The findings highlight the importance of attention mechanisms in understanding DNN vulnerabilities.