Multitask Learning Strengthens Adversarial Robustness

Mao, Chengzhi; Gupta, Amogh; Ray, Baishakhi; Song, Shuran; Yang, Junfeng; Vondrick, Carl

doi:10.48550/arxiv.2007.07236

Cited by 4 publications

(3 citation statements)

References 26 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Many defense methods against adversarial examples have been proposed to protect deep learning models [18][19][20][21][22][23][24][25].…”

Section: Defense Methods Against Adversarial Examplesmentioning

confidence: 99%

Random Transformation of image brightness for adversarial attack

Yang¹,

Xu²,

Wang³

et al. 2022

IFS

View full text Add to dashboard Cite

Deep neural networks (DNNs) are vulnerable to adversarial examples, which are crafted by adding small, human-imperceptible perturbations to the original images, but make the model output inaccurate predictions. Before DNNs are deployed, adversarial attacks can thus be an important method to evaluate and select robust models in safety-critical applications. However, under the challenging black-box setting, the attack success rate, i.e., the transferability of adversarial examples, still needs to be improved. Based on image augmentation methods, this paper found that random transformation of image brightness can eliminate overfitting in the generation of adversarial examples and improve their transferability. In light of this phenomenon, this paper proposes an adversarial example generation method, which can be integrated with Fast Gradient Sign Method (FGSM)-related methods to build a more robust gradient-based attack and to generate adversarial examples with better transferability. Extensive experiments on the ImageNet dataset have demonstrated the effectiveness of the aforementioned method. Whether on normally or adversarially trained networks, our method has a higher success rate for black-box attacks than other attack methods based on data augmentation. It is hoped that this method can help evaluate and improve the robustness of models.

show abstract

“…Many defense methods against adversarial examples have been proposed to protect deep learning models [18][19][20][21][22][23][24][25].…”

Section: Defense Methods Against Adversarial Examplesmentioning

confidence: 99%

Random Transformation of image brightness for adversarial attack

Yang¹,

Xu²,

Wang³

et al. 2022

IFS

View full text Add to dashboard Cite

show abstract

“…Multiple defense mechanisms have been proposed to protect deep learning models from the threat of adversarial examples [21][22][23][24][25][26][27][28]. Among these, adversarial training is the most effective way to improve model robustness [6,29] [30].…”

Section: Adversarial Defense Methodsmentioning

confidence: 99%

Improving the Transferability of Adversarial Examples with Image Affine Transformation

Yin

Zhang

et al. 2021

J. Phys.: Conf. Ser.

View full text Add to dashboard Cite

Deep learning is widely regarded as a black-box technology. We all know its performance is very good, but we have limited understanding of why it is so good. At present, there are many researches on the interpretability of deep neural network. By studying adversarial example, we can understand the internal semantics of neural network and find the decision boundary with problems, which in turn helps to improve the robustness and performance of neural network and its interpretability. With the development of adversarial example research, more and more adversarial example generation methods are proposed. Although the attack from adversarial example poses a great security threat to the deep learning system, it can also be used as an effective tool to measure the robustness and reliability of the model, and the attack and defense are two mutually promoting processes. Therefore, how to generate adversarial example with stronger attack ability is worth further study. And this study proposes a method named Affine-Invariant Method, aimed to improve the transferability of adversarial examples in black-box environment.

show abstract

“…Furthermore, learning features for multiple tasks can act as a regularizer, improving generalization. Mao et al [4] illustrated that multi-task learning improves adversarial robustness, which is critical for safety applications.…”

Section: Introductionmentioning

confidence: 99%

OmniDet: Surround View Cameras Based Multi-Task Visual Perception Network for Autonomous Driving

Kumar

Yogamani

Rashed³

et al. 2021

IEEE Robot. Autom. Lett.

View full text Add to dashboard Cite

Surround View fisheye cameras are commonly deployed in automated driving for 360°near-field sensing around the vehicle. This work presents a multi-task visual perception network on unrectified fisheye images to enable the vehicle to sense its surrounding environment. It consists of six primary tasks necessary for an autonomous driving system: depth estimation, visual odometry, semantic segmentation, motion segmentation, object detection, and lens soiling detection. We demonstrate that the jointly trained model performs better than the respective single task versions. Our multi-task model has a shared encoder providing a significant computational advantage and has synergized decoders where tasks support each other. We propose a novel camera geometry based adaptation mechanism to encode the fisheye distortion model both at training and inference. This was crucial to enable training on the WoodScape dataset, comprised of data from different parts of the world collected by 12 different cameras mounted on three different cars with different intrinsics and viewpoints. Given that bounding boxes is not a good representation for distorted fisheye images, we also extend object detection to use a polygon with non-uniformly sampled vertices. We additionally evaluate our model on standard automotive datasets, namely KITTI and Cityscapes. We obtain the state-of-the-art results on KITTI for depth estimation and pose estimation tasks and competitive performance on the other tasks. We perform extensive ablation studies on various architecture choices and task weighting methodologies. A short video at https://youtu.be/xbSjZ5OfPes provides qualitative results.

show abstract

Multitask Learning Strengthens Adversarial Robustness

Cited by 4 publications

References 26 publications

Random Transformation of image brightness for adversarial attack

Random Transformation of image brightness for adversarial attack

Improving the Transferability of Adversarial Examples with Image Affine Transformation

OmniDet: Surround View Cameras Based Multi-Task Visual Perception Network for Autonomous Driving

Contact Info

Product

Resources

About