OCCROB: Efficient SMT-Based Occlusion Robustness Verification of Deep Neural Networks

被引:5
|
作者
Guo, Xingwu [1 ]
Zhou, Ziwei [1 ]
Zhang, Yueling [1 ]
Katz, Guy [2 ]
Zhang, Min [1 ]
机构
[1] East China Normal Univ, Shanghai Key Lab Trustworthy Comp, Shanghai, Peoples R China
[2] Hebrew Univ Jerusalem, Jerusalem, Israel
关键词
D O I
10.1007/978-3-031-30823-9_11
中图分类号
TP31 [计算机软件];
学科分类号
081202 ; 0835 ;
摘要
Occlusion is a prevalent and easily realizable semantic perturbation to deep neural networks (DNNs). It can fool a DNN into misclassifying an input image by occluding some segments, possibly resulting in severe errors. Therefore, DNNs planted in safety-critical systems should be verified to be robust against occlusions prior to deployment. However, most existing robustness verification approaches for DNNs are focused on non-semantic perturbations and are not suited to the occlusion case. In this paper, we propose the first efficient, SMT-based approach for formally verifying the occlusion robustness of DNNs. We formulate the occlusion robustness verification problem and prove it is NP-complete. Then, we devise a novel approach for encoding occlusions as a part of neural networks and introduce two acceleration techniques so that the extended neural networks can be efficiently verified using off-the-shelf, SMT-based neural network verification tools. We implement our approach in a prototype called OccRob and extensively evaluate its performance on benchmark datasets with various occlusion variants. The experimental results demonstrate our approach's effectiveness and efficiency in verifying DNNs' robustness against various occlusions, and its ability to generate counterexamples when these DNNs are not robust.
引用
收藏
页码:208 / 226
页数:19
相关论文
共 50 条
  • [41] An efficient test method for noise robustness of deep neural networks
    Yasuda, Muneki
    Sakata, Hironori
    Cho, Seung-Il
    Harada, Tomochika
    Tanaka, Atushi
    Yokoyama, Michio
    IEICE NONLINEAR THEORY AND ITS APPLICATIONS, 2019, 10 (02): : 221 - 235
  • [42] Verification of Neural Networks' Global Robustness
    Kabaha, Anan
    Cohen, Dana Drachsler
    PROCEEDINGS OF THE ACM ON PROGRAMMING LANGUAGES-PACMPL, 2024, 8 (OOPSLA):
  • [43] Compositional Convolutional Neural Networks: A Deep Architecture with Innate Robustness to Partial Occlusion
    Kortylewski, Adam
    He, Ju
    Liu, Qing
    Yuille, Alan
    2020 IEEE/CVF CONFERENCE ON COMPUTER VISION AND PATTERN RECOGNITION (CVPR 2020), 2020, : 8937 - 8946
  • [44] Formal Modeling and SMT-Based Parameterized Verification of Data-Aware BPMN
    Calvanese, Diego
    Ghilardi, Silvio
    Gianola, Alessandro
    Montali, Marco
    Rivkin, Andrey
    BUSINESS PROCESS MANAGEMENT (BPM 2019), 2019, 11675 : 157 - 175
  • [45] ILP/SMT-Based Method for Design of Boolean Networks Based on Singleton Attractors
    Kobayashi, Koichi
    Hiraishi, Kunihiko
    IEEE-ACM TRANSACTIONS ON COMPUTATIONAL BIOLOGY AND BIOINFORMATICS, 2014, 11 (06) : 1253 - 1259
  • [46] SMT-based verification of data-aware processes: a model-theoretic approach
    Calvanese, Diego
    Ghilardi, Silvio
    Gianola, Alessandro
    Montali, Marco
    Rivkin, Andrey
    MATHEMATICAL STRUCTURES IN COMPUTER SCIENCE, 2020, 30 (03) : 271 - 313
  • [47] An MILP Encoding for Efficient Verification of Quantized Deep Neural Networks
    Mistry, Samvid
    Saha, Indranil
    Biswas, Swarnendu
    IEEE TRANSACTIONS ON COMPUTER-AIDED DESIGN OF INTEGRATED CIRCUITS AND SYSTEMS, 2022, 41 (11) : 4445 - 4456
  • [48] Efficient Weighted Model Integration via SMT-Based Predicate Abstraction
    Morettin, Paolo
    Passerini, Andrea
    Sebastiani, Roberto
    PROCEEDINGS OF THE TWENTY-SIXTH INTERNATIONAL JOINT CONFERENCE ON ARTIFICIAL INTELLIGENCE, 2017, : 720 - 728
  • [49] Survey on Robustness Verification of Feedforward Neural Networks and Recurrent Neural Networks
    Liu Y.
    Yang P.-F.
    Zhang L.-J.
    Wu Z.-L.
    Feng Y.
    Ruan Jian Xue Bao/Journal of Software, 2023, 34 (07): : 1 - 33
  • [50] A Unifying View on SMT-Based Software Verification (vol 60, pg 299, 2018)
    Beyer, Dirk
    Dangl, Matthias
    Wendler, Philipp
    JOURNAL OF AUTOMATED REASONING, 2021, 65 (03) : 461 - 461