Concolic Testing for Deep Neural Networks

Youcheng Sun; Min Wu; Wenjie Ruan; Xiaowei Huang; Marta Kwiatkowska; Daniel Kroening

doi:10.1145/3238147.3238172

Concolic Testing for Deep Neural Networks

Youcheng Sun, Min Wu, Wenjie Ruan, Xiaowei Huang, Marta Kwiatkowska, Daniel Kroening

School of Electronics, Electrical Engineering and Computer Science

Research output: Chapter in Book/Report/Conference proceeding › Conference contribution

198 Citations (Scopus)

432 Downloads (Pure)

Abstract

Concolic testing combines program execution and symbolic analysis to explore the execution paths of a software program. In this paper, we develop the first concolic testing approach for Deep Neural Networks (DNNs). More specifically, we utilise quantified linear arithmetic over rationals to express test requirements that have been studied in the literature, and then develop a coherent method to perform concolic testing with the aim of better coverage. Our experimental results show the effectiveness of the concolic testing approach in both achieving high coverage and finding adversarial examples.

Original language	English
Title of host publication	Proceedings of the 33rd ACM/IEEE International Conference on Automated Software Engineering
Publisher	Association for Computing Machinery
Pages	109-119
Number of pages	11
ISBN (Print)	978-1-4503-5937-5
DOIs	https://doi.org/10.1145/3238147.3238172
Publication status	Published - 03 Sept 2018

Access to Document

10.1145/3238147.3238172Licence: Unspecified

Concolic testing for deep neural networks
Copyright 2019 ACM. This work is made available online in accordance with the publisher’s policies. Please refer to any applicable terms of use of the publisher.
Accepted author manuscript, 657 KBLicence: Unspecified

Cite this

@inproceedings{d28e40738af245919ed10d4a5c476786,

title = "Concolic Testing for Deep Neural Networks",

abstract = "Concolic testing combines program execution and symbolic analysis to explore the execution paths of a software program. In this paper, we develop the first concolic testing approach for Deep Neural Networks (DNNs). More specifically, we utilise quantified linear arithmetic over rationals to express test requirements that have been studied in the literature, and then develop a coherent method to perform concolic testing with the aim of better coverage. Our experimental results show the effectiveness of the concolic testing approach in both achieving high coverage and finding adversarial examples.",

author = "Youcheng Sun and Min Wu and Wenjie Ruan and Xiaowei Huang and Marta Kwiatkowska and Daniel Kroening",

year = "2018",

month = sep,

day = "3",

doi = "10.1145/3238147.3238172",

language = "English",

isbn = "978-1-4503-5937-5",

pages = "109--119",

booktitle = "Proceedings of the 33rd ACM/IEEE International Conference on Automated Software Engineering",

publisher = "Association for Computing Machinery",

address = "United States",

}

TY - GEN

T1 - Concolic Testing for Deep Neural Networks

AU - Sun, Youcheng

AU - Wu, Min

AU - Ruan, Wenjie

AU - Huang, Xiaowei

AU - Kwiatkowska, Marta

AU - Kroening, Daniel

PY - 2018/9/3

Y1 - 2018/9/3

N2 - Concolic testing combines program execution and symbolic analysis to explore the execution paths of a software program. In this paper, we develop the first concolic testing approach for Deep Neural Networks (DNNs). More specifically, we utilise quantified linear arithmetic over rationals to express test requirements that have been studied in the literature, and then develop a coherent method to perform concolic testing with the aim of better coverage. Our experimental results show the effectiveness of the concolic testing approach in both achieving high coverage and finding adversarial examples.

AB - Concolic testing combines program execution and symbolic analysis to explore the execution paths of a software program. In this paper, we develop the first concolic testing approach for Deep Neural Networks (DNNs). More specifically, we utilise quantified linear arithmetic over rationals to express test requirements that have been studied in the literature, and then develop a coherent method to perform concolic testing with the aim of better coverage. Our experimental results show the effectiveness of the concolic testing approach in both achieving high coverage and finding adversarial examples.

U2 - 10.1145/3238147.3238172

DO - 10.1145/3238147.3238172

M3 - Conference contribution

SN - 978-1-4503-5937-5

SP - 109

EP - 119

BT - Proceedings of the 33rd ACM/IEEE International Conference on Automated Software Engineering

PB - Association for Computing Machinery

ER -

Concolic Testing for Deep Neural Networks

Abstract

Access to Document

Fingerprint

Cite this