The paper introduces the SAR-AE-SFP-Attack method to create real physics adversarial examples for SAR imagery recognition. It addresses challenges faced by existing methods and demonstrates significant improvements in attack efficiency across different models. The experiments conducted validate the effectiveness of the proposed approach, showcasing its potential for enhancing security and robustness in SAR systems.
The content delves into the generation of adversarial examples in Synthetic Aperture Radar (SAR) imagery recognition. It introduces a novel method, SAR-AE-SFP-Attack, that alters scattering feature parameters to create real physics adversarial examples. The approach significantly enhances attack efficiency on various model architectures, demonstrating transferability across different perspectives.
The paper discusses the susceptibility of deep neural network-based SAR target recognition models to adversarial attacks. It highlights the limitations of existing methods and proposes a new approach, SAR-AE-SFP-Attack, which optimizes scattering feature parameters to generate real physics adversarial examples. Experimental results show improved attack efficiency on CNN-based and Transformer-based models.
The study focuses on addressing vulnerabilities in Synthetic Aperture Radar (SAR) target recognition models through the generation of real physics adversarial examples using the SAR-AE-SFP-Attack method. By altering scattering feature parameters, this approach enhances attack efficiency on different model types, showcasing significant transferability effects.
Egy másik nyelvre
a forrásanyagból
arxiv.org
Mélyebb kérdések