766 Hits in 6.4 sec

Adversarial Noise Attacks of Deep Learning Architectures – Stability Analysis via Sparse Modeled Signals [article]

Yaniv Romano, Aviad Aberdam, Jeremias Sulam, Michael Elad
2019 arXiv   pre-print
In this paper we analyze the stability of state-of-the-art deep-learning classification machines to adversarial perturbations, where we assume that the signals belong to the (possibly multi-layer) sparse  ...  Our analysis links between the stability of the classification to noise and the underlying structure of the signal, quantified by the sparsity of its representation under a fixed dictionary.  ...  Our analysis exposes the ingredients of the data model governing the sensitivity to adversarial attacks, and clearly shows that the later pursuit (L-BP) is more robust.  ... 
arXiv:1805.11596v3 fatcat:taob5csjpvgwxjugmc575l3fsq

Special Issue on the Mathematical Foundations of Deep Learning in Imaging Science

Joan Bruna, Eldad Haber, Gitta Kutyniok, Thomas Pock, René Vidal
2020 Journal of Mathematical Imaging and Vision  
Deep learning methods have become an omnipresent and highly successful part of recent approaches in imaging and vision.  ...  Many mathematically inclined researchers have a strong desire to understand the theoretical reasons for the success of these approaches and to find relations between deep learning and mathematically well-established  ...  It is addressed in the article "Adversarial Noise Attacks of Deep Learning Architectures-Stability Analysis via Sparse Modeled Signals" by Romano et al..  ... 
doi:10.1007/s10851-020-00955-8 fatcat:n5kayejz4fcxxk5dvfit3gsbru

Stabilizing Deep Tomographic Reconstruction [article]

Weiwen Wu, Dianlin Hu, Wenxiang Cong, Hongming Shan, Shaoyu Wang, Chuang Niu, Pingkun Yan, Hengyong Yu, Varut Vardhanabhuti, Ge Wang
2021 arXiv   pre-print
In particular, the study shows that ACID-based reconstruction is resilient against adversarial attacks, superior to classic sparsity-regularized reconstruction alone, and eliminates the three kinds of  ...  Tomographic image reconstruction with deep learning is an emerging field, but a recent landmark study reveals that several deep reconstruction networks are unstable for computed tomography (CT) and magnetic  ...  The Lipschitz continuity assumption is useful to assess the convergence of a deep reconstruction algorithm. In our previous section IV.C, we have verified the BREN property for the data used in [6].  ... 
arXiv:2008.01846v5 fatcat:z4kyhdtj5re2xoepvhqziaklju

Snooping Attacks on Deep Reinforcement Learning [article]

Matthew Inkawhich, Yiran Chen, Hai Li
2020 arXiv   pre-print
Adversarial attacks have exposed a significant security vulnerability in state-of-the-art machine learning models. Among these models include deep reinforcement learning agents.  ...  of adversarial examples.  ...  Experiments and analysis Although previous work has shown that DRL agents are vulnerable to adversarial noise, to our knowledge it is not clear whether it is the adversarial nature of the noise or simply  ... 
arXiv:1905.11832v2 fatcat:u6rtp7vg35co7gc7jfeiw24psu

Distributed Machine Learning for Wireless Communication Networks: Techniques, Architectures, and Applications [article]

S. Hu, X. Chen, W. Ni, E. Hossain, X. Wang
2020 arXiv   pre-print
We also discuss the potential adversarial attacks faced by DML applications, and describe state-of-the-art countermeasures to preserve privacy and security.  ...  Distributed machine learning (DML) techniques, such as federated learning, partitioned learning, and distributed reinforcement learning, have been increasingly applied to wireless communications.  ...  The capability of DSVM against the attacks also depends on the network architecture and attack intensities.  ... 
arXiv:2012.01489v1 fatcat:pdauhq4xbbepvf26clhpqnc2ci

Deep learning-driven wireless communication for edge-cloud computing: opportunities and challenges

Huaming Wu, Xiangyi Li, Yingjun Deng
2020 Journal of Cloud Computing: Advances, Systems and Applications  
As a classic model of deep learning, autoencoder is widely used in the design paradigms of communication system models.  ...  We highlight the intuitions and key technologies of deep learning-driven wireless communication from the aspects of end-to-end communication, signal detection, channel estimation and compression sensing  ...  Training at different SNRs Up to now, it is still not clear which signal-to-noise (SNR) ratio the deep learning model should be trained on.  ... 
doi:10.1186/s13677-020-00168-9 fatcat:7n6r2pozgfb5rgfwyxoxpqxq3q

NAS-OoD: Neural Architecture Search for Out-of-Distribution Generalization [article]

Haoyue Bai, Fengwei Zhou, Lanqing Hong, Nanyang Ye, S.-H. Gary Chan, Zhenguo Li
2021 arXiv   pre-print
Recent advances on Out-of-Distribution (OoD) generalization reveal the robustness of deep learning models against distribution shifts.  ...  However, existing works focus on OoD algorithms, such as invariant risk minimization, domain generalization, or stable learning, without considering the influence of deep model architectures on OoD generalization  ...  Figure 6 .Figure 7 . 67 Temporal stability of search architecture.(Better viewed in the zoom-in mode) Statistical analysis of searched architectures on different datasets.  ... 
arXiv:2109.02038v1 fatcat:p2hzi5iegrfuvcnvhidjywpf2m

Table of Contents [EDICS]

2020 IEEE Transactions on Signal Processing  
Unser 4688 Model-Based Deep Learning for One-Bit Compressive Sensing . . . . . . . . . . . . . . . . . . . . . . . S. Khobahi and M.  ...  Zhou 5795 Sparse Bayesian Learning With Dynamic Filtering for Inference of Time-Varying Sparse Signals . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .  ...  Optimization Methods for Signal Processing Support Recovery for Sparse Signals With Unknown Non-Stationary Modulation . . . . . Y. Xie, M. B. Wakin, and G.  ... 
doi:10.1109/tsp.2020.3045363 fatcat:wcnvdcy3rvhblh7rtxfe6gz4re

Generative Adversarial Networks (GANs) in networking: A comprehensive survey & evaluation

Hojjat Navidan, Parisa Fard Moshiri, Mohammad Nabati, Reza Shahbazian, Seyed Ali Ghorashi, Vahid Shah-Mansouri, David Windridge
2021 Computer Networks  
Despite the recency of their conception, Generative Adversarial Networks (GANs) constitute an extensively researched machine learning sub-field for the creation of synthetic data through deep generative  ...  In this paper, we demonstrate how this branch of machine learning can benefit multiple aspects of computer and communication networks, including mobile networks, network analysis, internet of things, physical  ...  One form of wireless attack that targets this task is signal spoofing. In this type of attack, the adversary aims to impersonate a legitimate transmitter.  ... 
doi:10.1016/j.comnet.2021.108149 fatcat:4ekgil24ijha3evmzruez63tdq

Table of Contents

2020 IEEE Transactions on Signal Processing  
Wang 4959 Security-Enhanced Filter Design for Stochastic Systems under Malicious Attack via Smoothed Signal Model and Multiobjective Estimation Method .  ...  Xiao 5276 Model-Based Deep Learning for One-Bit Compressive Sensing . . . . . . . . . . . . . . . . . . . . . . . S. Khobahi and M.  ... 
doi:10.1109/tsp.2020.3042287 fatcat:nh7viihaozhd7li3txtadnx5ui

DeepRLS: A Recurrent Network Architecture with Least Squares Implicit Layers for Non-blind Image Deconvolution [article]

Iaroslav Koshelev, Daniil Selikhanovych, Stamatios Lefkimmiatis
2021 arXiv   pre-print
The first is that it implicitly models an effective image prior that can adequately characterize the set of natural images, while the second is that it recovers the corresponding maximum a posteriori (  ...  In this work, we study the problem of non-blind image deconvolution and propose a novel recurrent network architecture that leads to very competitive restoration results of high image quality.  ...  Motion blur kernel estimation via deep learning.  ... 
arXiv:2112.05505v1 fatcat:iqnb774wujap3d6zo4r5qcw2ge

A Survey on Adversarial Deep Learning Robustness in Medical Image Analysis

Kyriakos D. Apostolidis, George A. Papakostas
2021 Electronics  
The evolution of hardware has helped researchers to develop many powerful Deep Learning (DL) models to face numerous challenging problems.  ...  However, recent studies have shown that CNN models are vulnerable under adversarial attacks with imperceptible perturbations.  ...  Conflicts of Interest: The authors declare no conflict of interest.  ... 
doi:10.3390/electronics10172132 fatcat:xc32hvmiijcd3pgbuzbx7npqxy

Table of Contents

2020 IEEE Signal Processing Letters  
Cui 875 Adversarial Transfer Learning for Deep Learning Based Automatic Modulation Classification . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .  ...  Aazhang 1115 Efficient Tracking of Sparse Signals via an Earth Mover's Distance Dynamics Regularizer . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .  ...  Du 1844 Optimality Verification of Tensor Completion Model via Self-Validation . . . . . . . . . . C. Liu, H. Shan, T. Ma, and B.  ... 
doi:10.1109/lsp.2020.3040840 fatcat:ezrfzwo6tjbkfhohq2tgec4m6y

Manifold Regularization for Locally Stable Deep Neural Networks [article]

Charles Jin, Martin Rinard
2020 arXiv   pre-print
Our regularizers are based on a sparsification of the graph Laplacian which holds with high probability when the data is sparse in high dimensions, as is common in deep learning.  ...  Empirically, our networks exhibit stability in a diverse set of perturbation models, including ℓ_2, ℓ_∞, and Wasserstein-based perturbations; in particular, we achieve 40 against an adaptive PGD attack  ...  Empirically, our networks exhibit robustness against a variety of adversarial models implementing 2 , ∞ , and Wasserstein-based attacks.  ... 
arXiv:2003.04286v2 fatcat:7v5tuul45vgy7jiljtannwg6um

Data Hiding with Deep Learning: A Survey Unifying Digital Watermarking and Steganography [article]

Olivia Byrnes, Wendy La, Hu Wang, Congbo Ma, Minhui Xue, Qi Wu
2021 arXiv   pre-print
This survey summarises recent developments in deep learning techniques for data hiding for the purposes of watermarking and steganography, categorising them based on model architectures and noise injection  ...  Data hiding is the process of embedding information into a noise-tolerant signal such as a piece of audio, video, or image.  ...  NOISE INJECTION TECHNIQUES Aside from model architecture, deep learning-based data hiding techniques can also be classified based on noise injection methods.  ... 
arXiv:2107.09287v1 fatcat:2sqcyzv6t5ccdiffk5cmag7tya
« Previous Showing results 1 — 15 out of 766 results