DeepStellar: Model-based quantitative analysis of stateful deep learning systems

Xiaoning Du, Xiaofei Xie, Yi Li, Lei Ma, Yang Liu, Jianjun Zhao

研究成果: 著書/レポートタイプへの貢献会議での発言

5 引用 (Scopus)

抄録

Deep Learning (DL) has achieved tremendous success in many cutting-edge applications. However, the state-of-the-art DL systems still suffer from quality issues. While some recent progress has been made on the analysis of feed-forward DL systems, little study has been done on the Recurrent Neural Network (RNN)-based stateful DL systems, which are widely used in audio, natural languages and video processing, etc. In this paper, we initiate the very first step towards the quantitative analysis of RNN-based DL systems. We model RNN as an abstract state transition system to characterize its internal behaviors. Based on the abstract model, we design two trace similarity metrics and five coverage criteria which enable the quantitative analysis of RNNs. We further propose two algorithms powered by the quantitative measures for adversarial sample detection and coverage-guided test generation. We evaluate DeepStellar on four RNN-based systems covering image classification and automated speech recognition. The results demonstrate that the abstract model is useful in capturing the internal behaviors of RNNs, and confirm that (1) the similarity metrics could effectively capture the differences between samples even with very small perturbations (achieving 97% accuracy for detecting adversarial samples) and (2) the coverage criteria are useful in revealing erroneous behaviors (generating three times more adversarial samples than random testing and hundreds times more than the unrolling approach).

元の言語英語
ホスト出版物のタイトルESEC/FSE 2019 - Proceedings of the 2019 27th ACM Joint Meeting European Software Engineering Conference and Symposium on the Foundations of Software Engineering
編集者Sven Apel, Marlon Dumas, Alessandra Russo, Dietmar Pfahl
出版者Association for Computing Machinery, Inc
ページ477-487
ページ数11
ISBN(電子版)9781450355728
DOI
出版物ステータス出版済み - 8 12 2019
イベント27th ACM Joint European Software Engineering Conference and Symposium on the Foundations of Software Engineering, ESEC/FSE 2019 - Tallinn, エストニア
継続期間: 8 26 20198 30 2019

出版物シリーズ

名前ESEC/FSE 2019 - Proceedings of the 2019 27th ACM Joint Meeting European Software Engineering Conference and Symposium on the Foundations of Software Engineering

会議

会議27th ACM Joint European Software Engineering Conference and Symposium on the Foundations of Software Engineering, ESEC/FSE 2019
エストニア
Tallinn
期間8/26/198/30/19

Fingerprint

Recurrent neural networks
Learning systems
Chemical analysis
Image classification
Speech recognition
Deep learning
Testing
Processing

All Science Journal Classification (ASJC) codes

  • Artificial Intelligence
  • Software

これを引用

Du, X., Xie, X., Li, Y., Ma, L., Liu, Y., & Zhao, J. (2019). DeepStellar: Model-based quantitative analysis of stateful deep learning systems. : S. Apel, M. Dumas, A. Russo, & D. Pfahl (版), ESEC/FSE 2019 - Proceedings of the 2019 27th ACM Joint Meeting European Software Engineering Conference and Symposium on the Foundations of Software Engineering (pp. 477-487). (ESEC/FSE 2019 - Proceedings of the 2019 27th ACM Joint Meeting European Software Engineering Conference and Symposium on the Foundations of Software Engineering). Association for Computing Machinery, Inc. https://doi.org/10.1145/3338906.3338954

DeepStellar : Model-based quantitative analysis of stateful deep learning systems. / Du, Xiaoning; Xie, Xiaofei; Li, Yi; Ma, Lei; Liu, Yang; Zhao, Jianjun.

ESEC/FSE 2019 - Proceedings of the 2019 27th ACM Joint Meeting European Software Engineering Conference and Symposium on the Foundations of Software Engineering. 版 / Sven Apel; Marlon Dumas; Alessandra Russo; Dietmar Pfahl. Association for Computing Machinery, Inc, 2019. p. 477-487 (ESEC/FSE 2019 - Proceedings of the 2019 27th ACM Joint Meeting European Software Engineering Conference and Symposium on the Foundations of Software Engineering).

研究成果: 著書/レポートタイプへの貢献会議での発言

Du, X, Xie, X, Li, Y, Ma, L, Liu, Y & Zhao, J 2019, DeepStellar: Model-based quantitative analysis of stateful deep learning systems. : S Apel, M Dumas, A Russo & D Pfahl (版), ESEC/FSE 2019 - Proceedings of the 2019 27th ACM Joint Meeting European Software Engineering Conference and Symposium on the Foundations of Software Engineering. ESEC/FSE 2019 - Proceedings of the 2019 27th ACM Joint Meeting European Software Engineering Conference and Symposium on the Foundations of Software Engineering, Association for Computing Machinery, Inc, pp. 477-487, 27th ACM Joint European Software Engineering Conference and Symposium on the Foundations of Software Engineering, ESEC/FSE 2019, Tallinn, エストニア, 8/26/19. https://doi.org/10.1145/3338906.3338954
Du X, Xie X, Li Y, Ma L, Liu Y, Zhao J. DeepStellar: Model-based quantitative analysis of stateful deep learning systems. : Apel S, Dumas M, Russo A, Pfahl D, 編集者, ESEC/FSE 2019 - Proceedings of the 2019 27th ACM Joint Meeting European Software Engineering Conference and Symposium on the Foundations of Software Engineering. Association for Computing Machinery, Inc. 2019. p. 477-487. (ESEC/FSE 2019 - Proceedings of the 2019 27th ACM Joint Meeting European Software Engineering Conference and Symposium on the Foundations of Software Engineering). https://doi.org/10.1145/3338906.3338954
Du, Xiaoning ; Xie, Xiaofei ; Li, Yi ; Ma, Lei ; Liu, Yang ; Zhao, Jianjun. / DeepStellar : Model-based quantitative analysis of stateful deep learning systems. ESEC/FSE 2019 - Proceedings of the 2019 27th ACM Joint Meeting European Software Engineering Conference and Symposium on the Foundations of Software Engineering. 編集者 / Sven Apel ; Marlon Dumas ; Alessandra Russo ; Dietmar Pfahl. Association for Computing Machinery, Inc, 2019. pp. 477-487 (ESEC/FSE 2019 - Proceedings of the 2019 27th ACM Joint Meeting European Software Engineering Conference and Symposium on the Foundations of Software Engineering).
@inproceedings{731e8255761e44939077b878c13fc3c5,
title = "DeepStellar: Model-based quantitative analysis of stateful deep learning systems",
abstract = "Deep Learning (DL) has achieved tremendous success in many cutting-edge applications. However, the state-of-the-art DL systems still suffer from quality issues. While some recent progress has been made on the analysis of feed-forward DL systems, little study has been done on the Recurrent Neural Network (RNN)-based stateful DL systems, which are widely used in audio, natural languages and video processing, etc. In this paper, we initiate the very first step towards the quantitative analysis of RNN-based DL systems. We model RNN as an abstract state transition system to characterize its internal behaviors. Based on the abstract model, we design two trace similarity metrics and five coverage criteria which enable the quantitative analysis of RNNs. We further propose two algorithms powered by the quantitative measures for adversarial sample detection and coverage-guided test generation. We evaluate DeepStellar on four RNN-based systems covering image classification and automated speech recognition. The results demonstrate that the abstract model is useful in capturing the internal behaviors of RNNs, and confirm that (1) the similarity metrics could effectively capture the differences between samples even with very small perturbations (achieving 97{\%} accuracy for detecting adversarial samples) and (2) the coverage criteria are useful in revealing erroneous behaviors (generating three times more adversarial samples than random testing and hundreds times more than the unrolling approach).",
author = "Xiaoning Du and Xiaofei Xie and Yi Li and Lei Ma and Yang Liu and Jianjun Zhao",
year = "2019",
month = "8",
day = "12",
doi = "10.1145/3338906.3338954",
language = "English",
series = "ESEC/FSE 2019 - Proceedings of the 2019 27th ACM Joint Meeting European Software Engineering Conference and Symposium on the Foundations of Software Engineering",
publisher = "Association for Computing Machinery, Inc",
pages = "477--487",
editor = "Sven Apel and Marlon Dumas and Alessandra Russo and Dietmar Pfahl",
booktitle = "ESEC/FSE 2019 - Proceedings of the 2019 27th ACM Joint Meeting European Software Engineering Conference and Symposium on the Foundations of Software Engineering",

}

TY - GEN

T1 - DeepStellar

T2 - Model-based quantitative analysis of stateful deep learning systems

AU - Du, Xiaoning

AU - Xie, Xiaofei

AU - Li, Yi

AU - Ma, Lei

AU - Liu, Yang

AU - Zhao, Jianjun

PY - 2019/8/12

Y1 - 2019/8/12

N2 - Deep Learning (DL) has achieved tremendous success in many cutting-edge applications. However, the state-of-the-art DL systems still suffer from quality issues. While some recent progress has been made on the analysis of feed-forward DL systems, little study has been done on the Recurrent Neural Network (RNN)-based stateful DL systems, which are widely used in audio, natural languages and video processing, etc. In this paper, we initiate the very first step towards the quantitative analysis of RNN-based DL systems. We model RNN as an abstract state transition system to characterize its internal behaviors. Based on the abstract model, we design two trace similarity metrics and five coverage criteria which enable the quantitative analysis of RNNs. We further propose two algorithms powered by the quantitative measures for adversarial sample detection and coverage-guided test generation. We evaluate DeepStellar on four RNN-based systems covering image classification and automated speech recognition. The results demonstrate that the abstract model is useful in capturing the internal behaviors of RNNs, and confirm that (1) the similarity metrics could effectively capture the differences between samples even with very small perturbations (achieving 97% accuracy for detecting adversarial samples) and (2) the coverage criteria are useful in revealing erroneous behaviors (generating three times more adversarial samples than random testing and hundreds times more than the unrolling approach).

AB - Deep Learning (DL) has achieved tremendous success in many cutting-edge applications. However, the state-of-the-art DL systems still suffer from quality issues. While some recent progress has been made on the analysis of feed-forward DL systems, little study has been done on the Recurrent Neural Network (RNN)-based stateful DL systems, which are widely used in audio, natural languages and video processing, etc. In this paper, we initiate the very first step towards the quantitative analysis of RNN-based DL systems. We model RNN as an abstract state transition system to characterize its internal behaviors. Based on the abstract model, we design two trace similarity metrics and five coverage criteria which enable the quantitative analysis of RNNs. We further propose two algorithms powered by the quantitative measures for adversarial sample detection and coverage-guided test generation. We evaluate DeepStellar on four RNN-based systems covering image classification and automated speech recognition. The results demonstrate that the abstract model is useful in capturing the internal behaviors of RNNs, and confirm that (1) the similarity metrics could effectively capture the differences between samples even with very small perturbations (achieving 97% accuracy for detecting adversarial samples) and (2) the coverage criteria are useful in revealing erroneous behaviors (generating three times more adversarial samples than random testing and hundreds times more than the unrolling approach).

UR - http://www.scopus.com/inward/record.url?scp=85070657609&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=85070657609&partnerID=8YFLogxK

U2 - 10.1145/3338906.3338954

DO - 10.1145/3338906.3338954

M3 - Conference contribution

AN - SCOPUS:85070657609

T3 - ESEC/FSE 2019 - Proceedings of the 2019 27th ACM Joint Meeting European Software Engineering Conference and Symposium on the Foundations of Software Engineering

SP - 477

EP - 487

BT - ESEC/FSE 2019 - Proceedings of the 2019 27th ACM Joint Meeting European Software Engineering Conference and Symposium on the Foundations of Software Engineering

A2 - Apel, Sven

A2 - Dumas, Marlon

A2 - Russo, Alessandra

A2 - Pfahl, Dietmar

PB - Association for Computing Machinery, Inc

ER -