Loading [MathJax]/extensions/MathMenu.js
DeepManeuver: Adversarial Test Generation for Trajectory Manipulation of Autonomous Vehicles | IEEE Journals & Magazine | IEEE Xplore

DeepManeuver: Adversarial Test Generation for Trajectory Manipulation of Autonomous Vehicles


Abstract:

Adversarial test generation techniques aim to produce input perturbations that cause a DNN to compute incorrect outputs. For autonomous vehicles driven by a DNN, however,...Show More

Abstract:

Adversarial test generation techniques aim to produce input perturbations that cause a DNN to compute incorrect outputs. For autonomous vehicles driven by a DNN, however, the effect of such perturbations are attenuated by other parts of the system and are less effective as vehicle state evolves. In this work we argue that for adversarial testing perturbations to be effective on autonomous vehicles, they must account for the subtle interplay between the DNN and vehicle states. Building on that insight, we develop DeepManeuver, an automated framework that interleaves adversarial test generation with vehicle trajectory physics simulation. Thus, as the vehicle moves along a trajectory, DeepManeuver enables the refinement of candidate perturbations to: (1) account for changes in the state of the vehicle that may affect how the perturbation is perceived by the system; (2) retain the effect of the perturbation on previous states so that the current state is still reachable and past trajectory is preserved; and (3) result in multi-target maneuvers that require fulfillment of vehicle state sequences (e.g. reaching locations in a road to navigate a tight turn). Our assessment reveals that DeepManeuver can generate perturbations to force maneuvers more effectively and consistently than state-of-the-art techniques by 20.7 percentage points on average. We also show DeepManeuver's effectiveness at disrupting vehicle behavior to achieve multi-target maneuvers with a minimum 52% rate of success.
Published in: IEEE Transactions on Software Engineering ( Volume: 49, Issue: 10, 01 October 2023)
Page(s): 4496 - 4509
Date of Publication: 08 August 2023

ISSN Information:

Funding Agency:

University of Virginia, Charlottesville, VA, USA
Meriel von Stein is a Ph.D. candidate and founding member of the LESS Lab at the University of Virginia. She is an Organizing Committee Member of the inaugural SE4SafeML 2023 workshop colocated with the Foundations of Software Engineering (FSE) conference. She was the recipient of the University of Virginia Graduate Teaching Award and serves as the Mentoring Chair of the UVA Computer Science Graduate Student Group. Her re...Show More
Meriel von Stein is a Ph.D. candidate and founding member of the LESS Lab at the University of Virginia. She is an Organizing Committee Member of the inaugural SE4SafeML 2023 workshop colocated with the Foundations of Software Engineering (FSE) conference. She was the recipient of the University of Virginia Graduate Teaching Award and serves as the Mentoring Chair of the UVA Computer Science Graduate Student Group. Her re...View more
University of Virginia, Charlottesville, VA, USA
AI Division, Software Engineering Institute, Pittsburgh, PA, USA
David Shriver received the Ph.D. degree from the University of Virginia, in December 2022. Since contributing to this article, he has been working as a Machine Learning Research Scientist with the AI Division at the Software Engineering Institute. He received the 2023 ACM SIGSOFT Outstanding Doctoral Dissertation Award for the thesis “Increasing the Applicability of Verification Tools for Neural Networks.” His research in...Show More
David Shriver received the Ph.D. degree from the University of Virginia, in December 2022. Since contributing to this article, he has been working as a Machine Learning Research Scientist with the AI Division at the Software Engineering Institute. He received the 2023 ACM SIGSOFT Outstanding Doctoral Dissertation Award for the thesis “Increasing the Applicability of Verification Tools for Neural Networks.” His research in...View more
University of Virginia, Charlottesville, VA, USA
Sebastian Elbaum (Fellow, IEEE) is the Anita Jones Professor with the Department of Computer Science at the University of Virginia, where he Co-Leads the Laboratory for Engineering Safe Software (LESS Lab). He was the recipient of an NSF Career Award, an IBM Innovation Award, a Google Faculty Research Award, an FSE Test of Time Award, five ACM SigSoft Distinguished Paper Awards, and multiple best paper awards. He regularl...Show More
Sebastian Elbaum (Fellow, IEEE) is the Anita Jones Professor with the Department of Computer Science at the University of Virginia, where he Co-Leads the Laboratory for Engineering Safe Software (LESS Lab). He was the recipient of an NSF Career Award, an IBM Innovation Award, a Google Faculty Research Award, an FSE Test of Time Award, five ACM SigSoft Distinguished Paper Awards, and multiple best paper awards. He regularl...View more

University of Virginia, Charlottesville, VA, USA
Meriel von Stein is a Ph.D. candidate and founding member of the LESS Lab at the University of Virginia. She is an Organizing Committee Member of the inaugural SE4SafeML 2023 workshop colocated with the Foundations of Software Engineering (FSE) conference. She was the recipient of the University of Virginia Graduate Teaching Award and serves as the Mentoring Chair of the UVA Computer Science Graduate Student Group. Her research focuses on developing testing techniques for robotic systems with machine learned components. Her recent work includes adversarial testing for perception subsystems and distribution-aware input repair for perception subsystems.
Meriel von Stein is a Ph.D. candidate and founding member of the LESS Lab at the University of Virginia. She is an Organizing Committee Member of the inaugural SE4SafeML 2023 workshop colocated with the Foundations of Software Engineering (FSE) conference. She was the recipient of the University of Virginia Graduate Teaching Award and serves as the Mentoring Chair of the UVA Computer Science Graduate Student Group. Her research focuses on developing testing techniques for robotic systems with machine learned components. Her recent work includes adversarial testing for perception subsystems and distribution-aware input repair for perception subsystems.View more
University of Virginia, Charlottesville, VA, USA
AI Division, Software Engineering Institute, Pittsburgh, PA, USA
David Shriver received the Ph.D. degree from the University of Virginia, in December 2022. Since contributing to this article, he has been working as a Machine Learning Research Scientist with the AI Division at the Software Engineering Institute. He received the 2023 ACM SIGSOFT Outstanding Doctoral Dissertation Award for the thesis “Increasing the Applicability of Verification Tools for Neural Networks.” His research interests include analysis and verification of neural networks, especially in the context of systems with a high cost of failure.
David Shriver received the Ph.D. degree from the University of Virginia, in December 2022. Since contributing to this article, he has been working as a Machine Learning Research Scientist with the AI Division at the Software Engineering Institute. He received the 2023 ACM SIGSOFT Outstanding Doctoral Dissertation Award for the thesis “Increasing the Applicability of Verification Tools for Neural Networks.” His research interests include analysis and verification of neural networks, especially in the context of systems with a high cost of failure.View more
University of Virginia, Charlottesville, VA, USA
Sebastian Elbaum (Fellow, IEEE) is the Anita Jones Professor with the Department of Computer Science at the University of Virginia, where he Co-Leads the Laboratory for Engineering Safe Software (LESS Lab). He was the recipient of an NSF Career Award, an IBM Innovation Award, a Google Faculty Research Award, an FSE Test of Time Award, five ACM SigSoft Distinguished Paper Awards, and multiple best paper awards. He regularly serves in program committees at the top software engineering and robotic conferences, and has served as a Program Co-Chair for ISSTA07, ESEM08, and ICSE2015, and as the Steering Committee Chair for ICSE. He is an ACM fellow. His latest work focuses on robotic systems with learned components. His research aims to build dependable systems through domain-specific analysis techniques.
Sebastian Elbaum (Fellow, IEEE) is the Anita Jones Professor with the Department of Computer Science at the University of Virginia, where he Co-Leads the Laboratory for Engineering Safe Software (LESS Lab). He was the recipient of an NSF Career Award, an IBM Innovation Award, a Google Faculty Research Award, an FSE Test of Time Award, five ACM SigSoft Distinguished Paper Awards, and multiple best paper awards. He regularly serves in program committees at the top software engineering and robotic conferences, and has served as a Program Co-Chair for ISSTA07, ESEM08, and ICSE2015, and as the Steering Committee Chair for ICSE. He is an ACM fellow. His latest work focuses on robotic systems with learned components. His research aims to build dependable systems through domain-specific analysis techniques.View more

Contact IEEE to Subscribe

References

References is not available for this document.