• Home
  • Search
  • Towards robust image classification with deep learning and real-time DNN inference on mobile
  • Cite Icon2
  • https://doi.org/10.17760/d20416558Copy DOI Icon

Towards robust image classification with deep learning and real-time DNN inference on mobile

  • Jan 1, 2020
  • Pu Zhao
Show More
  • Abstract
  • Literature Map
  • References
  • Citations
  • Similar Papers
Abstract

As the rapidly increasing popularity of deep learning, deep neural networks (DNN) have become the fundamental and essential building blocks in various applications such as image classification and object detection. However, there are two main issues which potentially limit the wide application of DNNs: 1) the robustness of DNN models raises security concerns, and 2) the large computation and storage requirements of DNN models lead to difficulties for its wide deployment on popular yet resource-constrained devices such as mobile phones. To investigate the DNN robustness, we explore the DNN attack, robustness evaluation and defense. More specifically, for DNN attack, we achieve various attack goals (e.g. adversarial examples and fault sneaking attacks) with different algorithms (e.g. alternating direction method of multipliers (ADMM) and natural gradient descent (NGD) attacks) under various conditions (white-box and black-box attacks). For robustness evaluation, we propose a fast evaluation method to obtain the model perturbation bound such that any model perturbation within the bound does not alter the model classification outputs or incur model mis-behaviors. For the DNN defense, we investigate the defense performance with model connection techniques and successfully mitigate the fault sneaking and backdoor attacks. With a deeper understanding of the DNN robustness, we further explore the deployment problem of DNN models on edge devices with limited resources. To satisfy the storage and computation limitation on edge devices, we adopt model pruning to remove the redundancy in models, thus reducing the storage and computation during inference. Besides, as some applications have real-time requirements with high inference speed sensitivities such as object detection on autonomous cars, we further try to implement real-time DNN inference for various DNN applications on mobile devices with pruning and compiler optimization. To summary, we mainly investigate the DNN robustness and implement real-time DNN inference on the mobile.--Author's abstract

Similar Papers
  • Research Article
  • Citations8

ACT-Detector: Adaptive channel transformation-based light-weighted detector for adversarial attacks

  • Feb 16, 2021
  • Information Sciences
  • Jinyin Chen +4
  • Research Article

Transfer-based Adversarial Attack with Rectified Adam and Color Invariance

  • Jan 01, 2022
  • International Journal of Software and Informatics
  • Jia Ding +1
  • PDF
  • Research Article
  • Citations67

Assessing Optimizer Impact on DNN Model Sensitivity to Adversarial Examples

  • Jan 01, 2019
  • IEEE Access
  • Yixiang Wang +5
  • Conference Article
  • Citations4

Defending Deep Learning Based Anomaly Detection Systems Against White-Box Adversarial Examples and Backdoor Attacks

  • Nov 12, 2020
  • Khaled Alrawashdeh +1
  • Dissertation

High-performance and energy-efficient deep learning for resource-constrained devices

  • Jan 01, 2020
  • Ao Ren
  • Research Article
  • Citations37

Fooling deep neural detection networks with adaptive object-oriented adversarial perturbation

  • Feb 20, 2021
  • Pattern Recognition
  • Yatie Xiao +2
  • Research Article
  • Citations6

Timing Black-Box Attacks: Crafting Adversarial Examples through Timing Leaks against DNNs on Embedded Devices

  • Jul 09, 2021
  • IACR Transactions on Cryptographic Hardware and Embedded Systems
  • Tsunato Nakai +2
  • Research Article
  • Citations9

Invisible backdoor attack with attention and steganography

  • Oct 19, 2024
  • Computer Vision and Image Understanding
  • Wenmin Chen +5
  • Research Article
  • Citations6

Adversarial-Inspired Backdoor Defense via Bridging Backdoor and Adversarial Attacks

  • Apr 11, 2025
  • Proceedings of the AAAI Conference on Artificial Intelligence
  • Jia-Li Yin +4
  • Research Article
  • Citations2

Restoration of Adversarial Examples Using Image Arithmetic Operations

  • Jan 01, 2022
  • Intelligent Automation & Soft Computing
  • Kazim Ali +1
  • Conference Article
  • Citations4177

Practical Black-Box Attacks against Machine Learning

  • Apr 02, 2017
  • Nicolas Papernot +5
  • Conference Article
  • Citations6

Edge devices object detection by filter pruning

  • Sep 07, 2021
  • Viviana Crescitelli +3
  • Research Article
  • Citations23

IWA: Integrated gradient‐based white‐box attacks for fooling deep neural networks

  • Oct 28, 2021
  • International Journal of Intelligent Systems
  • Yixiang Wang +4
  • Research Article
  • Citations28

PArtNNer: Platform-Agnostic Adaptive Edge-Cloud DNN Partitioning for Minimizing End-to-End Latency

  • Jan 10, 2024
  • ACM Transactions on Embedded Computing Systems
  • Soumendu Kumar Ghosh +3
  • PDF
  • Research Article
  • Citations33

Boosting Targeted Black-Box Attacks via Ensemble Substitute Training and Linear Augmentation

  • Jun 03, 2019
  • Applied Sciences
  • Xianfeng Gao +4
Cactus Communications logo

Copyright 2026 Cactus Communications. All rights reserved.