Multiclass ASMA vs Targeted PGD Attack in Image Segmentation
This highlights a vulnerability in deep learning networks for image segmentation, posing risks for applications relying on such models, though it is incremental as it compares existing attacks on standard architectures.
The paper investigated the effectiveness of the projected gradient descent (PGD) attack and Adaptive Mask Segmentation Attack (ASMA) on the DeepLabV3 image segmentation model using MobileNetV3 and ResNet50 architectures, finding that PGD consistently altered segmentation to its target while ASMA's generalization to multiclass targets was less effective.
Deep learning networks have demonstrated high performance in a large variety of applications, such as image classification, speech recognition, and natural language processing. However, there exists a major vulnerability exploited by the use of adversarial attacks. An adversarial attack imputes images by altering the input image very slightly, making it nearly undetectable to the naked eye, but results in a very different classification by the network. This paper explores the projected gradient descent (PGD) attack and the Adaptive Mask Segmentation Attack (ASMA) on the image segmentation DeepLabV3 model using two types of architectures: MobileNetV3 and ResNet50, It was found that PGD was very consistent in changing the segmentation to be its target while the generalization of ASMA to a multiclass target was not as effective. The existence of such attack however puts all of image classification deep learning networks in danger of exploitation.