A Self-supervised Approach for Adversarial Robustness

By Muzammal Naseer, Salman Hameed Khan, Munawar Hayat, Fahad Shahbaz Khan, Fatih Murat Porikli. Computer Vision and Pattern Recognition, 2020

371 citations recorded by the source corpus.

Adversarial examples can cause catastrophic mistakes in Deep Neural Network (DNNs) based vision systems e.g., for classification, segmentation and object detection. The vulnerability of DNNs against such attacks can prove a major roadblock towards their real-world deployment. Transferability of adversarial examples demand generalizable defenses that can provide cross-task protection. Adversarial training that enhances robustness by modifying target model's parameters lacks such generalizability. On the other hand, different input processing based defenses fall short in the face of continuously evolving attacks. In this paper, we take the first step to combine the benefits of both approaches and propose a self-supervised adversarial training mechanism in the input space. By design, our defense is a generalizable approach and provides significant robustness against the \textbf{unseen} adversarial attacks (\eg by reducing the success rate of translation-invariant \textbf{ensemble} attack from 82.6\% to 31.9\% in comparison to previous state-of-the-art). It can be deployed as a plug-and-play solution to protect a variety of vision systems, as we demonstrate for the case of classification, segmentation and detection.

Subfields: Artificial Intelligence, Electrical and Electronic Engineering

View Paper: https://www.semanticscholar.org/paper/2c0564ca47262f76a46bfacc6d587a968be53712

How researchers see this paper

The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.

In Artificial Intelligence

1. Hang Su
2. Jun Zhu
3. Tianyu Pang
4. Yinpeng Dong
5. Ajmal Saeed Mian
6. Mubarak Shah
7. Xiao Yang
8. Naveed Akhtar

In Electrical and Electronic Engineering

1. Tianyu Pang
2. Jun Zhu
3. Shuicheng Yan
4. Mubarak Shah
5. Hang Su
6. Xiao Yang
7. Yinpeng Dong
8. Sung Ju Hwang

Citations in the corpus

• Advances in adversarial attacks and defenses in computer vision: A survey (2021, positive): “[341] proposed self-supervised adversarial training, whereas adversarial training is independently analyzed for self-supervision by incorporating it in pretraining in [342].”
• Better Diffusion Models Further Improve Adversarial Training (2023, background): “, 2020a;b), self-supervised learning (Chen et al., 2020a;b; Naseer et al., 2020; Wang & Liu, 2022), ensemble learning (Tramèr et al.”
• Enhancing the Transferability of Adversarial Attacks through Variance Tuning (2021, methodology): “[23] design a neural representation purifier (NRP) model that learns to purify the adversarially perturbed images based on the automatically derived supervision.”
• Adversarial Self-Supervised Contrastive Learning (2020, background): “Using self-supervision signal in terms of perceptual loss also shows effective results in denoising the adversarial perturbation as purifier network [34].”
• Bag of Tricks for Adversarial Training (2020, background): “…al., 2020c), generative modeling (Jiang et al., 2018; Pang et al., 2018b; Wang & Yu, 2019; Deng et al., 2020), semi-supervised learning (Carmon et al., 2019; Alayrac et al., 2019; Zhai et al., 2019), and self-supervised learning (Hendrycks et al., 2019; Chen et al., 2020a;b; Naseer et al., 2020).”