Denoised Internal Models: A Brain-inspired Autoencoder Against Adversarial Attacks

doi:10.1007/s11633-022-1375-7

	Denoised Internal Models: A Brain-inspired Autoencoder Against Adversarial Attacks
	Liu, Kai-Yuan 1,3; Li, Xing-Yu 2; Lai, Yu-Rui1 ; Su, Hang1 ; Wang, Jia-Chen1 ; Guo, Chun-Xu 1; Xie, Hong 5,6; Guan, Ji-Song1 ; Zhou, Yi4
	2022-10
发表期刊	MACHINE INTELLIGENCE RESEARCH (IF:6.4[JCR-2023],6.4[5-Year])
ISSN	2731-538X
EISSN	2731-5398
卷号	19 期号:5 页码:456-471
DOI	10.1007/s11633-022-1375-7
摘要	Despite its great success, deep learning severely suffers from robustness; i.e., deep neural networks are very vulnerable to adversarial attacks, even the simplest ones. Inspired by recent advances in brain science, we propose the denoised internal models (DIM), a novel generative autoencoder-based model to tackle this challenge. Simulating the pipeline in the human brain for visual signal processing, DIM adopts a two-stage approach. In the first stage, DIM uses a denoiser to reduce the noise and the dimensions of inputs, reflecting the information pre-processing in the thalamus. Inspired by the sparse coding of memory-related traces in the primary visual cortex, the second stage produces a set of internal models, one for each category. We evaluate DIM over 42 adversarial attacks, showing that DIM effectively defenses against all the attacks and outperforms the SOTA on the overall robustness on the MNIST (Modified National Institute of Standards and Technology) dataset. © 2022, Institute of Automation, Chinese Academy of Sciences and Springer-Verlag GmbH Germany, part of Springer Nature.
关键词	Brain Learning systems Long short-term memory Pipeline processing systems Signal processing Adversarial attack Auto encoders Brain science Brain-inspired Brain-inspired learning Generative model Human brain Internal models Robustness Simple++
URL	查看原文
收录类别	EI ; ESCI
语种	英语
出版者	Chinese Academy of Sciences
EI入藏号	20224112860287
EI主题词	Deep neural networks
EI分类号	461.1 Biomedical Engineering ; 461.4 Ergonomics and Human Factors Engineering ; 716.1 Information Theory and Signal Processing ; 722.4 Digital Computers and Systems
原始文献类型	Journal article (JA)
文献类型	期刊论文
条目标识符	https://kms.shanghaitech.edu.cn/handle/2MSLDSTB/241086
专题	信息科学与技术学院_硕士生信息科学与技术学院_外聘教师生命科学与技术学院_PI研究组_管吉松组信息科学与技术学院_本科生信息科学与技术学院_博士生
通讯作者	Guan, Ji-Song; Zhou, Yi
作者单位	1.School of Life Sciences and Technology, ShanghaiTech University, Shanghai; 201210, China; 2.Shanghai Center for Brain Science and Brain-inspired Technology, Shanghai; 201602, China; 3.School of Life Sciences, Tsinghua University, Beijing; 100084, China; 4.National Engineering Laboratory for Brain-inspired Intelligence Technology and Application, School of Information Science and Technology, University of Science and Technology of China, Hefei; 230026, China; 5.Institute of Photonic Chips, University of Shanghai for Science and Technology, Shanghai; 200093, China; 6.Centre for Artificial-intelligence Nanophotonics, School of Optical-electrical and Computer Engineering, University of Shanghai for Science and Technology, Shanghai; 200093, China
第一作者单位	上海科技大学
通讯作者单位	上海科技大学
第一作者的第一单位	上海科技大学
推荐引用方式 GB/T 7714	Liu, Kai-Yuan,Li, Xing-Yu,Lai, Yu-Rui,et al. Denoised Internal Models: A Brain-inspired Autoencoder Against Adversarial Attacks[J]. MACHINE INTELLIGENCE RESEARCH,2022,19(5):456-471.
APA	Liu, Kai-Yuan.,Li, Xing-Yu.,Lai, Yu-Rui.,Su, Hang.,Wang, Jia-Chen.,...&Zhou, Yi.(2022).Denoised Internal Models: A Brain-inspired Autoencoder Against Adversarial Attacks.MACHINE INTELLIGENCE RESEARCH,19(5),456-471.
MLA	Liu, Kai-Yuan,et al."Denoised Internal Models: A Brain-inspired Autoencoder Against Adversarial Attacks".MACHINE INTELLIGENCE RESEARCH 19.5(2022):456-471.