About

Matthew B. Dwyer is the Robert Thomson Distinguished Professor at the Laboratory for Engineering Safe Software within the Department of Computer Science at the University of Virginia. His research focuses on engineering safe software, which involves developing methods and tools to ensure software reliability, security, and correctness. As a distinguished professor, he has made significant contributions to the field of computer science, particularly in areas related to software safety and engineering.

Research topics

• Computer Science
• Artificial Intelligence
• Machine Learning
• Computer Security
• Data Mining
• Distributed computing
• Programming language

Selected publications

• Input Distribution Coverage: Measuring Feature Interaction Adequacy in Neural Network Testing

  ACM Transactions on Software Engineering and Methodology · 2022 · 20 citations

  • Computer Science
  • Computer Science
  • Machine Learning

  Testing deep neural networks (DNNs) has garnered great interest in the recent years due to their use in many applications. Black-box test adequacy measures are useful for guiding the testing process in covering the input domain. However, the absence of input specifications makes it challenging to apply black-box test adequacy measures in DNN testing. The Input Distribution Coverage (IDC) framework addresses this challenge by using a variational autoencoder to learn a low dimensional latent representation of the input distribution, and then using that latent space as a coverage domain for testing. IDC applies combinatorial interaction testing on a partitioning of the latent space to measure test adequacy. Empirical evaluation demonstrates that IDC is cost-effective, capable of detecting feature diversity in test inputs, and more sensitive than prior work to test inputs generated using different DNN test generation methods. The findings demonstrate that IDC overcomes several limitations of white-box DNN coverage approaches by discounting coverage from unrealistic inputs and enabling the calculation of test adequacy metrics that capture the feature diversity present in the input space of DNNs.

  PublisherOA PDFDOI

• Reducing DNN Properties to Enable Falsification with Adversarial Attacks

  2021 · 18 citations

  Senior authorCorresponding
  • Computer Science
  • Computer Science
  • Artificial Intelligence

  Deep Neural Networks (DNN) are increasingly being deployed in safety-critical domains, from autonomous vehicles to medical devices, where the consequences of errors demand techniques that can provide stronger guarantees about behavior than just high test accuracy. This paper explores broadening the application of existing adversarial attack techniques for the falsification of DNN safety properties. We contend and later show that such attacks provide a powerful repertoire of scalable algorithms for property falsification. To enable the broad application of falsification, we introduce a semantics-preserving reduction of multiple safety property types, which subsume prior work, into a set of equivalid correctness problems amenable to adversarial attacks. We evaluate our reduction approach as an enabler of falsification on a range of DNN correctness problems and show its cost-effectiveness and scalability.

  DOI

Recent grants

• SHF: Small: Distribution-aware Testing for Neural Networks

  NSF · $498k · 2021–2025

• SHF: Small: Measurable Program Analysis

  NSF · $228k · 2018–2020

• Collaborative Research: Finite-State Verification for High-Performance Computing

  NSF · $300k · 2006–2010

• SHF: Medium: Rearchitecting Neural Networks for Verification

  NSF · $1.3M · 2019–2024

• SHF: Small: Measurable Program Analysis

  NSF · $508k · 2016–2018

Frequent coauthors

• Sebastian Elbaum

  University of Virginia

  55 shared

• John Hatcliff

  Kansas State University

  49 shared

• Robby

  22 shared

• Cleon Anderson

  Parsons (United States)

  21 shared

• Corina S. Păsăreanu

  19 shared

• Willem Visser

  Stellenbosch University

  16 shared

• James C. Corbett

  Lawrence Livermore National Laboratory

  15 shared

• Myra B. Cohen

  Iowa State University

  13 shared

Labs

• LESS LabPI

Education

• Doctor of Philosophy, Computer Science

  University of Massachusetts Amherst

  1995

• Masters in Computer Science, Computer Science

  University of Massachusetts Boston

  1989

• Bachelors of Science in Electrical Engineering, Electrical and Computer Engineering

  University of Rochester

  1985

Awards & honors

• SIGSOFT ISSTA Test of Time Award 2022
• IEEE Computer Society Harlan D. Mills Award 2022
• SIGSOFT Impact Paper Award 2021
• SIGSOFT Distinguished Service Award 2019
• LERO David Lorge Parnas Fellow 2018

Similar researchers at University of Virginia

• Stephen S. Richh-170
• Gary Owensh-107
• Jarrod Martoh-101
• Jennifer L. Westh-93
• Edward H. Egelmanh-93