702 Hits in 10.1 sec

I Know What You See: Power Side-Channel Attack on Convolutional Neural Network Accelerators [article]

Lingxiao Wei, Yannan Liu, Bo Luo, Yu Li, Qiang Xu
2018 arXiv   pre-print
To be specific, we perform the attack on an FPGA-based convolutional neural network accelerator and we manage to recover the input image from the collected power traces without knowing the detailed parameters  ...  For the MNIST dataset, our power side-channel attack is able to achieve up to 89% recognition accuracy.  ...  In this paper, we present a power side-channel attack on an FPGA-based convolutional neural network (CNN) accelerator.  ... 
arXiv:1803.05847v1 fatcat:jkatht4vf5cvnig4yssqzcnfju

Detecting Dead Weights and Units in Neural Networks [article]

Utku Evci
2018 arXiv   pre-print
Deep Neural Networks are highly over-parameterized and the size of the neural networks can be reduced significantly after training without any decrease in performance.  ...  One can clearly see this phenomenon in a wide range of architectures trained for various problems.  ...  [1] focuses on pruning bayesian neural networks. [9] shows that stochastic quantization helps preventing black box attacks on neural networks.  ... 
arXiv:1806.06068v1 fatcat:hba7ejd5drda5iwoqbvuesdl7q

Connections between Numerical Algorithms for PDEs and Neural Networks [article]

Tobias Alt, Karl Schrader, Matthias Augustin, Pascal Peter, Joachim Weickert
2021 arXiv   pre-print
We connect these concepts to residual networks, recurrent neural networks, and U-net architectures.  ...  Our goal is to transfer the rich set of mathematical foundations from the world of PDEs to neural networks.  ...  They often suffer from undesirable side effects such as a sensitivity against adversarial attacks [37] .  ... 
arXiv:2107.14742v1 fatcat:unlaensgzrhahobkyho6qs6pnm

Edge Intelligence: Architectures, Challenges, and Applications [article]

Dianlei Xu, Tong Li, Yong Li, Xiang Su, Sasu Tarkoma, Tao Jiang, Jon Crowcroft, Pan Hui
2020 arXiv   pre-print
In this paper, we present a thorough and comprehensive survey on the literature surrounding edge intelligence.  ...  Edge intelligence refers to a set of connected systems and devices for data collection, caching, processing, and analysis in locations close to where data is captured based on artificial intelligence.  ...  However, there is a side effect brought on by group convolution, i.e., outputs of one channel are only derived from a small part of the input channels.  ... 
arXiv:2003.12172v2 fatcat:xbrylsvb7bey5idirunacux6pe

Cyberthreats in 2025

James Bret Michael, Richard Kuhn, Jeffrey Voas
2020 Computer  
GARFINKEL: Adversarial images, like putting a sticker on a stop sign to make a convolutional neural network think that it's a "speed limit 55" sign, makes for a popular research topic because the results  ...  New AI-based cybersecurity simulate attacks on the networks to constantly fix the holes in the networks and predict the attacks.  ... 
doi:10.1109/mc.2020.2983529 fatcat:35vcw7pdjvg7xlmaxviqdrg6ga

Construction of Value Chain E-Commerce Model Based on Stationary Wavelet Domain Deep Residual Convolutional Neural Network

Chenyuan Wang, Wei Wang
2020 Complexity  
By training the optimal parameters of the deep residual network and comparing the results with other models, the method of this paper has a good effect against the sample.  ...  The original loss function based on the residual learning model deep learning is modified to solve the original model fuzzy problem, which improves the effect and has good robustness.  ...  Researches on image enhancement and noise processing based on convolutional neural networks have proposed related methods [30, 31] .  ... 
doi:10.1155/2020/6611325 fatcat:uo6dbf3rhne35f3apna5exvv7a

Automated Side Channel Analysis of Media Software with Manifold Learning [article]

Yuanyuan Yuan, Qi Pang, Shuai Wang
2021 arXiv   pre-print
This paper explores an adversary's ability to launch side channel analyses (SCA) against media software to reconstruct confidential media inputs.  ...  be addressed in a unified manner with an autoencoder framework trained to learn the mapping between media inputs and side channel observations.  ...  The encoder φ θ comprises several stacked 2D convolutional neural networks (CNNs).  ... 
arXiv:2112.04947v2 fatcat:sfuail7w3faptgvufnaroqwdc4

Learning with Vertically-Partitioned Data, Binary Feedback, and Random Parameter Update

Ngu Nguyen, Stephan Sigg
2019 IEEE INFOCOM 2019 - IEEE Conference on Computer Communications Workshops (INFOCOM WKSHPS)  
For instance, the authentication question "What have you done today?" or "What have you not done after 11:00am?" is shown above the images.  ...  Hence, such passwords suffer from vulnerability to shoulder-surfing [49] and side-channel attacks [12] .  ...  An auto-encoder [156] is an unsupervised learning technique in which a feedforward non-recurrent neural network is trained to reproduce an input.  ... 
doi:10.1109/infcomw.2019.8845203 dblp:conf/infocom/NguyenS19 fatcat:cjgyjbtyhrejhdn37okuaq3nmi

Physics-based Deep Learning [article]

Nils Thuerey and Philipp Holl and Maximilian Mueller and Patrick Schnell and Felix Trost and Kiwon Um
2021 arXiv   pre-print
As much as possible, all topics come with hands-on code examples in the form of Jupyter notebooks to quickly get started.  ...  We live in exciting times: these methods have a huge potential to fundamentally change what computer simulations can achieve.  ...  Because of this it’s important to work on getting to know the data you are dealing with.  ... 
arXiv:2109.05237v2 fatcat:dm2wyckg6fcxzhsxi4hmo76sny

Cellular, Wide-Area, and Non-Terrestrial IoT: A Survey on 5G Advances and the Road Towards 6G [article]

Mojtaba Vaezi, Amin Azari, Saeed R. Khosravirad, Mahyar Shirvanimoghaddam, M. Mahdi Azari, Danai Chasaki, Petar Popovski
2021 arXiv   pre-print
The solutions include grant-free access and channel coding for short-packet communications, non-orthogonal multiple access, and on-device intelligence.  ...  This paper provides a comprehensive survey on existing and emerging communication solutions for serving IoT applications in the context of cellular, wide-area, as well as non-terrestrial networks.  ...  Gbps (see Table I and Fig. 4) .  ... 
arXiv:2107.03059v1 fatcat:i7rxlipsd5eojgpx3x5yxioozq

Applications of Deep Neural Networks [article]

Jeff Heaton
2021 arXiv   pre-print
This course will introduce the student to classic neural network structures, Convolution Neural Networks (CNN), Long Short-Term Memory (LSTM), Gated Recurrent Neural Networks (GRU), General Adversarial  ...  It is not necessary to know Python prior to this book; however, familiarity with at least one programming language is assumed.  ...  Typically you will either see 1 color channel (grayscale) or 3 color channels (RGB color). If we look inside of one of the 50,000 elements we can see the structure of each image.  ... 
arXiv:2009.05673v3 fatcat:jlvjj475xrd6dap7izuuxvl4le

Kinetic Song Comprehension: Deciphering Personal Listening Habits via Phone Vibrations [article]

Richard Matovu, Isaac Griswold-Steiner, Abdul Serwadda
2019 arXiv   pre-print
This paper demonstrates a new way in which motion sensor data can be leveraged to intrude on user music preferences without their express permission.  ...  Although users can mitigate some of the risk by using a phone cover to dampen the vibrations, we show that a sophisticated attacker could adapt the attack to still classify songs with a decent accuracy  ...  These results point to motion sensors as a powerful side-channel for leakage of information on the music which a smartphone user listens to. 2) Evaluating defensive technique and attacker countermeasures  ... 
arXiv:1909.09123v1 fatcat:nl3vbfefcbgqfiwelxuhmu67cm

Serdab: An IoT Framework for Partitioning Neural Networks Computation across Multiple Enclaves [article]

Tarek Elgamal, Klara Nahrstedt
2020 arXiv   pre-print
Our partitioning strategy achieves up to 4.7x speedup compared to executing the entire neural network in one enclave.  ...  To bridge this gap, Serdab presents a DNN partitioning strategy to distribute the layers of the neural network across multiple enclave devices or across an enclave device and other hardware accelerators  ...  We thank Mikhail Tadjikov, John Maguire, and Ingrid Guch who are members of The Aerospace corporation for their valuable feedback on this work.  ... 
arXiv:2005.06043v1 fatcat:cp4kdpo6pngnpdy2pyesmolvnq


2021 2021 IEEE International Conference on Consumer Electronics-Taiwan (ICCE-TW)  
Deep Neural Networks (DNN) accelerators.  ...  convolutional neural networks.  ... 
doi:10.1109/icce-tw52618.2021.9602919 fatcat:aetmvxb7hfah7iuucbamos2wgu

How You Act Tells a Lot: Privacy-Leakage Attack on Deep Reinforcement Learning [article]

Xinlei Pan, Weiyao Wang, Xiaoshuai Zhang, Bo Li, Jinfeng Yi, Dawn Song
2019 arXiv   pre-print
However, most of these studies focus on supervised learning models.  ...  To explore such privacy breaches in general, we mainly propose two methods: environment dynamics search via genetic algorithm and candidate inference based on shadow policies.  ...  WHAT MAP DID YOU WALK ON?  ... 
arXiv:1904.11082v1 fatcat:7uh3gjuxmfgv3if676dhfxdvp4
« Previous Showing results 1 — 15 out of 702 results