Towards an Effective and Interpretable Refinement Approach for DNN Verification

Loading...
Thumbnail Image
File version

Accepted Manuscript (AM)

Author(s)
Li, J
Bai, G
Pham, LH
Sun, J
Griffith University Author(s)
Primary Supervisor
Other Supervisors
Editor(s)
Date
2023
Size
File type(s)
Location

Chiang Mai, Thailand

License
Abstract

Recently, several abstraction refinement techniques have been proposed to improve the verification precision for deep neural networks (DNNs). However, these techniques usually take many refinement steps to verify a property and the refinement decision in each step is hard to interpret, thus hindering their analysis, reasoning and optimization.In this work, we propose SURGEON, a novel DNN verification refinement approach that is both effective and interpretable, allowing analyst to understand why and how each refinement decision is made. The main insight is to leverage the 'interpretable' nature of debugging processes and formulate the verification refinement problem as a debugging problem. Given a failed verification procedure, SURGEON refines it in an iterative manner and, in each iteration, it effectively identifies the root cause of the failure and heuristically generates fixes according to abstract transformers.We have implemented SURGEON in a prototype and evaluated it using a set of local robustness verification problems. Besides the interpretability, the experimental results show our approach can improve the precision of base verification methods and is more effective than existing refinement techniques.

Journal Title
Conference Title

2023 IEEE 23rd International Conference on Software Quality, Reliability, and Security (QRS)

Book Title
Edition
Volume
Issue
Thesis Type
Degree Program
School
Publisher link
Patent number
Funder(s)
Grant identifier(s)
Rights Statement
Rights Statement

This work is covered by copyright. You must assume that re-use is limited to personal use and that permission from the copyright owner must be obtained for all other uses. If the document is available under a specified licence, refer to the licence for details of permitted re-use. If you believe that this work infringes copyright please make a copyright takedown request using the form at https://www.griffith.edu.au/copyright-matters.

Item Access Status
Note
Access the data
Related item(s)
Subject

Quantum engineering systems (incl. computing and communications)

Software quality, processes and metrics

Neural networks

Persistent link to this record
Citation

Li, J; Bai, G; Pham, LH; Sun, J, Towards an Effective and Interpretable Refinement Approach for DNN Verification, 2023 IEEE 23rd International Conference on Software Quality, Reliability, and Security (QRS), 2023, pp. 569-580