Adversarial attacks pose a significant threat to deep learning models, specifically medical images, as they can mislead models into making inaccurate predictions by introducing subtle distortions to the input data that are often imperceptible to humans.
Abstract: Deep-Iearning-based fault diagnosis methods have been proved effective in recent years. With lots of convolutional-neural-network-based models ...
People also ask
What are adversarial attacks in deep learning?
What method to enhance deep learning fault diagnosis by generating adversarial samples?
What are adversarial attacks on AI models?
What is the difference between PGD and BIM attack?
Abstract—Deep-learning-based fault diagnosis methods have been proved effective in recent years. With lots of convolutional- neural-network-based models ...
The results show that when either white-box or black-box adversarial attacks are applied the models are vulnerable and fail in detecting faults and can help ...
People also search for
In particular, the work of (Paschali et al., 2018) tested whether existing medical deep learning models can be attacked by adversarial attacks. They showed ...
Missing: fault | Show results with:fault
Mar 21, 2024 · This study explores the threats in deploying deep learning models for fault diagnosis in ACS using the Tennessee Eastman Process dataset. By ...
May 18, 2024 · This study explores the threats in deploying deep learning models for Fault Detection and Diagnosis (FDD) in ACS using the Tennessee Eastman ...
While machine learning models have shown superior performance in fault diagnosis systems, researchers have revealed their vulnerability to subtle noises ...
Dec 11, 2020 · The aim of the study was to test the feasibility and impact of an adversarial attack on the accuracy of a deep learning-based dermatoscopic ...
Missing: fault | Show results with:fault
The objective of adversarial attacks targeting classification tasks is to strategically modify input images to elicit inaccurate classification. The decision- ...
Missing: fault | Show results with:fault