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

Qi Liu, Tao Liu, Zihao Liu, Yanzhi Wang, Yier Jin, Wujie Wen

Research output: Chapter in Book/Entry/PoemConference contribution

37 Scopus citations

Abstract

Thanks to recent machine learning model innovation and computing hardware advancement, the state-of-the-art of Deep Neural Network (DNN) is presenting human-level performance for many complex intelligent tasks in real-world applications. However, it also introduces ever-increasing security concerns for those intelligent systems. For example, the emerging adversarial attacks indicate that even very small and often imperceptible adversarial input perturbations can easily mislead the cognitive function of deep learning systems (DLS). Existing DNN adversarial studies are narrowly performed on the ideal software-level DNN models with a focus on single uncertainty factor, i.e. input perturbations, however, the impact of DNN model reshaping on adversarial attacks, which is introduced by various hardware-favorable techniques such as hash-based weight compression during modern DNN hardware implementation, has never been discussed. In this work, we for the first time investigate the multi-factor adversarial attack problem in practical model optimized deep learning systems by jointly considering the DNN model-reshaping (e.g. HashNet based deep compression) and the input perturbations. We first augment adversarial example generating method dedicated to the compressed DNN models by incorporating the software-based approaches and mathematical modeled DNN reshaping. We then conduct a comprehensive robustness and vulnerability analysis of deep compressed DNN models under derived adversarial attacks. A defense technique named 'gradient inhibition' is further developed to ease the generating of adversarial examples thus to effectively mitigate adversarial attacks towards both software and hardware-oriented DNNs. Simulation results show that 'gradient inhibition' can decrease the average success rate of adversarial attacks from 87.99% to 4.77% (from 86.74% to 4.64%) on MNIST (CIFAR-10) benchmark with marginal accuracy degradation across various DNNs.

Original languageEnglish (US)
Title of host publicationASP-DAC 2018 - 23rd Asia and South Pacific Design Automation Conference, Proceedings
PublisherInstitute of Electrical and Electronics Engineers Inc.
Pages721-726
Number of pages6
ISBN (Electronic)9781509006021
DOIs
StatePublished - Feb 20 2018
Event23rd Asia and South Pacific Design Automation Conference, ASP-DAC 2018 - Jeju, Korea, Republic of
Duration: Jan 22 2018Jan 25 2018

Publication series

NameProceedings of the Asia and South Pacific Design Automation Conference, ASP-DAC
Volume2018-January

Other

Other23rd Asia and South Pacific Design Automation Conference, ASP-DAC 2018
Country/TerritoryKorea, Republic of
CityJeju
Period1/22/181/25/18

ASJC Scopus subject areas

  • Electrical and Electronic Engineering
  • Computer Science Applications
  • Computer Graphics and Computer-Aided Design

Fingerprint

Dive into the research topics of 'Security analysis and enhancement of model compressed deep learning systems under adversarial attacks'. Together they form a unique fingerprint.

Cite this