MARBLE: model-based robustness analysis of stateful deep learning systems

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

Research output: Chapter in Book/Report/Conference proceedingConference PaperResearchpeer-review

Abstract

State-of-the-art deep learning (DL) systems are vulnerable to adversarial examples, which hinders their potential adoption in safety-and security-critical scenarios. While some recent progress has been made in analyzing the robustness of feed-forward neural networks, the robustness analysis for stateful DL systems, such as recurrent neural networks (RNNs), still remains largely uncharted. In this paper, we propose Marble, a model-based approach for quantitative robustness analysis of real-world RNN-based DL systems. Marble builds a probabilistic model to compactly characterize the robustness of RNNs through abstraction. Furthermore, we propose an iterative refinement algorithm to derive a precise abstraction, which enables accurate quantification of the robustness measurement. We evaluate the effectiveness of Marble on both LSTM and GRU models trained separately with three popular natural language datasets. The results demonstrate that (1) our refinement algorithm is more efficient in deriving an accurate abstraction than the random strategy, and (2) Marble enables quantitative robustness analysis, in rendering better efficiency, accuracy, and scalability than the state-of-the-art techniques.

Original languageEnglish
Title of host publicationProceedings - 2020 35th IEEE/ACM International Conference on Automated Software Engineering, ASE 2020
EditorsClaire Le Goues, David Lo
Place of PublicationNew York NY USA
PublisherIEEE, Institute of Electrical and Electronics Engineers
Pages423-435
Number of pages13
ISBN (Electronic)9781450367684
DOIs
Publication statusPublished - 2020
Externally publishedYes
EventAutomated Software Engineering Conference 2020 - Virtual, Melbourne, Australia
Duration: 21 Sep 202025 Sep 2020
Conference number: 35th
https://dl.acm.org/doi/proceedings/10.1145/3324884 (Proceedings)
https://conf.researchr.org/home/ase-2020 (Website)
https://dl.acm.org/doi/proceedings/10.1145/3417113 (Proceedings)

Conference

ConferenceAutomated Software Engineering Conference 2020
Abbreviated titleASE 2020
CountryAustralia
CityMelbourne
Period21/09/2025/09/20
Internet address

Cite this