Input-Relational Verification of Deep Neural Networks

被引:1
|
作者
Banerjee, Debangshu [1 ]
Xu, Changming [1 ]
Singh, Gagandeep [1 ,2 ]
机构
[1] Univ Illinois, Urbana, IL 61801 USA
[2] VMware Res, Palo Alto, CA USA
来源
PROCEEDINGS OF THE ACM ON PROGRAMMING LANGUAGES-PACMPL | 2024年 / 8卷 / PLDI期
关键词
Abstract Interpretation; Deep Learning; Relational Verification; ABSTRACT DOMAIN;
D O I
10.1145/3656377
中图分类号
TP31 [计算机软件];
学科分类号
081202 ; 0835 ;
摘要
We consider the verification of input-relational properties defined over deep neural networks (DNNs) such as robustness against universal adversarial perturbations, monotonicity, etc. Precise verification of these properties requires reasoning about multiple executions of the same DNN. We introduce a novel concept of difference tracking to compute the difference between the outputs of two executions of the same DNN at all layers. We design a new abstract domain, DiffPoly for efficient difference tracking that can scale large DNNs. DiffPoly is equipped with custom abstract transformers for common activation functions (ReLU, Tanh, Sigmoid, etc.) and affine layers and can create precise linear cross-execution constraints. We implement an input-relational verifier for DNNs called RaVeN which uses DiffPoly and linear program formulations to handle a wide range of input-relational properties. Our experimental results on challenging benchmarks show that by leveraging precise linear constraints defined over multiple executions of the DNN, RaVeN gains substantial precision over baselines on a wide range of datasets, networks, and input-relational properties.
引用
收藏
页数:27
相关论文
共 50 条
  • [21] Detecting unusual input to neural networks
    Martin, Joerg
    Elster, Clemens
    APPLIED INTELLIGENCE, 2021, 51 (04) : 2198 - 2209
  • [22] Formal verification for quantized neural networks
    Kovasznai, Gergely
    Kiss, Dorina Hedvig
    Mlinko, Peter
    ANNALES MATHEMATICAE ET INFORMATICAE, 2023, 57 : 36 - 48
  • [23] Formal verification for quantized neural networks
    Kovasznai, Gergely
    Kiss, Dorina Hedvig
    Mlinko, Peter
    ANNALES MATHEMATICAE ET INFORMATICAE, 2023, 57 : 36 - 48
  • [24] Automating Visual Inspection of Lyophilized Drug Products With Multi-Input Deep Neural Networks
    Tsay, Calvin
    Li, Zheng
    2019 IEEE 15TH INTERNATIONAL CONFERENCE ON AUTOMATION SCIENCE AND ENGINEERING (CASE), 2019, : 1802 - 1807
  • [25] Ranking with Deep Neural Networks
    Prakash, Chandan
    Sarkar, Amitrajit
    PROCEEDINGS OF 2018 FIFTH INTERNATIONAL CONFERENCE ON EMERGING APPLICATIONS OF INFORMATION TECHNOLOGY (EAIT), 2018,
  • [26] On the Singularity in Deep Neural Networks
    Nitta, Tohru
    NEURAL INFORMATION PROCESSING, ICONIP 2016, PT IV, 2016, 9950 : 389 - 396
  • [27] Orthogonal Deep Neural Networks
    Li, Shuai
    Jia, Kui
    Wen, Yuxin
    Liu, Tongliang
    Tao, Dacheng
    IEEE TRANSACTIONS ON PATTERN ANALYSIS AND MACHINE INTELLIGENCE, 2021, 43 (04) : 1352 - 1368
  • [28] Relational Turkish text classification with graph neural networks
    Okur, Halil Ibrahim
    Tohma, Kadir
    Sertbas, Ahmet
    JOURNAL OF POLYTECHNIC-POLITEKNIK DERGISI, 2024,
  • [29] Deep Morphological Neural Networks
    Shen, Yucong
    Shih, Frank Y.
    Zhong, Xin
    Chang, I-Cheng
    INTERNATIONAL JOURNAL OF PATTERN RECOGNITION AND ARTIFICIAL INTELLIGENCE, 2022, 36 (12)
  • [30] A new approach to modeling cycles with summer and winter demand peaks as input variables for deep neural networks
    Jasinski, Tomasz
    RENEWABLE & SUSTAINABLE ENERGY REVIEWS, 2022, 159