• Home
  • Search
  • Towards robust and secure deep learning models and beyond
  • Cite Icon1
  • https://doi.org/10.17760/d20467286Copy DOI Icon

Towards robust and secure deep learning models and beyond

  • Jan 1, 2022
  • Siyue Wang
Show More
  • Abstract
  • Literature Map
  • References
  • Citations
  • Similar Papers
Abstract

Modern science and technology witness the breakthroughs of deep learning during the past decades. Fueled by the rapid improvements of computational resources, learning algorithms, and massive amount of data, deep neural networks (DNNs) have played a dominant role in more and more real world applications. Nonetheless, there is a spring of bitterness mingling with this remarkable success - recent studies have revealed the limitations of DNNs which raise safety and reliability concerns of its widespread usage: 1) the robustness of DNN models under adversarial attacks and the instability problem on edge devices, and 2) the protection and verification of intellectual properties of well-trained DNN models. In this thesis, we first focus on investigating and tackling the problem of how to build robust DNNs under adversarial attacks. In adversarial attacks, deliberately crafted small perturbations added to the clean input can lead to wrong prediction results with high confidence. We approach the solution by incorporating stochasticity into DNN models. We propose multiple schemes to harden the DNN models when facing adversarial threats, including Defensive Dropout (DD), Hierarchical Random Switching (HRS), and Adversarially Trained Model Switching (AdvMS). Apart from that, we also propose a stochastic fault-tolerant training scheme that can generally improve the robustness of DNNs when facing the instability problem on DNN accelerators without focusing on optimizations for individual devices. The second part of this thesis focuses on how to effectively protect the intellectual property for DNNs and reliably identify their ownership. We propose Characteristic Examples (C-examples) for effectively fingerprinting DNN models, featuring high-robustness to the well-trained DNN and its derived versions (e.g. pruned models) as well as low-transferability to unassociated models. To better perform functionality verification of DNNs implemented on edge devices for on-device inference applications, we also propose Intrinsic Examples. Intrinsic Examples as fingerprinting of DNN can also detect adversarial third-party attacks that embed misbehaviors through re-training. The generation process of our fingerprints does not intervene with the training phase and no additional data are required from the training/testing set.--Author's abstract

Similar Papers
  • Conference Article
  • Citations5

Investigating the Robustness of Deep Neural Network Based COVID-19 Detection Models Against Universal Adversarial Attacks

  • Dec 18, 2021
  • Mohammad Akidul Hoque +3
  • Conference Article
  • Citations33

DNNGuard: An Elastic Heterogeneous DNN Accelerator Architecture against Adversarial Attacks

  • Mar 09, 2020
  • Xingbin Wang +6
  • Conference Article
  • Citations23

Security analysis and enhancement of model compressed deep learning systems under adversarial attacks

  • Jan 22, 2018
  • arXiv (Cornell University)
  • Qi Liu +5
  • Research Article
  • Citations13

SecureNet: Proactive intellectual property protection and model security defense for DNNs based on backdoor learning

  • Feb 21, 2024
  • Neural Networks
  • Peihao Li +4
  • Conference Article
  • Citations12

Is Approximation Universally Defensive Against Adversarial Attacks in Deep Neural Networks?

  • Mar 14, 2022
  • Ayesha Siddique +1
  • Conference Article
  • Citations6

Mitigating Adversarial Attacks for Deep Neural Networks by Input Deformation and Augmentation

  • Jan 01, 2020
  • Pengfei Qiu +5
  • Conference Article
  • Citations34

Model Reverse-Engineering Attack using Correlation Power Analysis against Systolic Array Based Neural Network Accelerator

  • Oct 01, 2020
  • Kota Yoshida +4
  • Research Article
  • Citations3

AccelAT: A Framework for Accelerating the Adversarial Training of Deep Neural Networks Through Accuracy Gradient

  • Jan 01, 2022
  • IEEE Access
  • Farzad Nikfam +3
  • Conference Article
  • Citations2

An Empirical Study on Robustness of DNNs with Out-of-Distribution Awareness

  • Dec 01, 2020
  • Lingjun Zhou +6
  • Conference Article
  • Citations1

Understanding adversarial attack and defense towards deep compressed neural networks

  • May 03, 2018
  • Qi Liu +2
  • Conference Article
  • Citations4

Fast Robustness Prediction for Deep Neural Network

  • Oct 28, 2019
  • Yuehuan Wang +4
  • Research Article
  • Citations5

SSAT: Active Authorization Control and User’s Fingerprint Tracking Framework for DNN IP Protection

  • Oct 29, 2024
  • ACM Transactions on Multimedia Computing, Communications, and Applications
  • Mingfu Xue +5
  • Research Article
  • Citations7

ActiveGuard: An active intellectual property protection technique for deep neural networks by leveraging adversarial examples as users' fingerprints

  • May 10, 2023
  • IET Computers & Digital Techniques
  • Mingfu Xue +6
  • Research Article
  • Citations7

Predicted Robustness as QoS for Deep Neural Network Models

  • Sep 30, 2020
  • Journal of Computer Science and Technology
  • Yue-Huan Wang +5
  • Conference Article
  • Citations1

Enhancing the Robustness of Deep Neural Networks from "Smart" Compression

  • Jul 01, 2018
  • Tao Liu +3
Cactus Communications logo

Copyright 2026 Cactus Communications. All rights reserved.