Natural Images Allow Universal Adversarial Attacks on Medical Image Classification Using Deep Neural Networks with Transfer Learning

Natural Images Allow Universal Adversarial Attacks on Medical Image Classification Using Deep Neural Networks with Transfer Learning

Transfer learning from natural images is used in deep neural networks (DNNs) for medical image classification to achieve a computer-aided clinical diagnosis. Although the adversarial vulnerability of DNNs hinders practical applications owing to the high stakes of diagnosis, adversarial attacks are e...

Full description

Bibliographic Details
Main Authors: Akinori Minagi, Hokuto Hirano, Kauzhiro Takemoto
Format: Article
Language:English
Published: MDPI AG 2022-02-01
Series:Journal of Imaging
Subjects:
deep neural networks
transfer learning
medical imaging
adversarial attacks
security
privacy
Online Access:https://www.mdpi.com/2313-433X/8/2/38

Internet

https://www.mdpi.com/2313-433X/8/2/38

Similar Items