Is Robustness the Cost of Accuracy? – A Comprehensive Study on the Robustness of 18 Deep Image Classification Models

Abstract

The prediction accuracy has been the long-lasting and sole standard for comparing the performance of different image classification models, including the ImageNet competition. However, recent studies have highlighted the lack of robustness in well-trained deep neural networks to adversarial examples. Visually imperceptible perturbations to natural images can easily be crafted and mislead the image classifiers towards misclassification. To demystify the trade-offs between robustness and accuracy, in this paper we thoroughly benchmark 18 ImageNet models using multiple robustness metrics, including the distortion, success rate and transferability of adversarial examples between 306 pairs of models. Our extensive experimental results reveal several new insights:(1) linear scaling law -the empirical 2 and ∞ distortion metrics scale linearly with the logarithm of classification error; (2) model architecture is a more critical factor to robustness than model size, and the disclosed accuracy-robustness Pareto frontier can be used as an evaluation criterion for ImageNet model designers; (3) for a similar network architecture, increasing network depth slightly improves robustness in ∞ distortion; (4) there exist models (in VGG family) that exhibit high adversarial transferability, while most adversarial examples crafted from one model can only be transferred within the same family. Experiment code is publicly available at https://github.com/huanzhang12/Adversarial Survey.where f (x, t) is a loss function to measure the distance between the prediction of x and the target label t. In this work, we choose f (x, t) = max{max i =t [(Logit(x)) i − (Logit(x)) t ], −κ}

Adversarial Attacks and Defenses in Images, Graphs and Text: A Review

“…It is important to note that x may not necessarily follow the distribution D. Thus, the studies on adversarial examples are different from these on model generalization. Moreover, a number of studies reported the relation between these two properties Su et al, 2018;Stutz et al, 2019;Zhang et al, 2019b). From our clarification, we hope that our audience get the difference and relation between risk and adversarial risk, and the importance of studying adversarial countermeasures.…”

Adversarial explanations for understanding image classification decisions and improved neural network robustness

“…A number of stochastic and non-differentiable defenses have been proposed and subsequently shown to be vulnerable to attacks which take these qualities into account [9]. Still other defense papers have focused on defense against specifically single-step attacks [21], were marginally less effective versions of the previously mentioned, state-of-the-art approaches [22], or focused on the natural defensive qualities of different architectures rather than ways of improving their defenses [23].…”

AdvMS: A Multi-Source Multi-Cost Defense Against Adversarial Attacks

“…In this method, although one can strengthen the defending effectiveness by using adversarial examples with large distortions, it also leads to degraded classification accuracy on natural images. In fact, although the ultimate goal of defenders is to design effective defense with negligible harms on other factors, some researchers also point out the trade-off between adversarial robustness and its cost factor (such as test accuracy) may be an inevitable nature of deep neural nets [21,22].…”