CommPact: Evaluating the Feasibility of Autonomous Vehicle Contracts

Dec 1, 2018·
Jeremy Erickson
Shibo Chen
Shibo Chen
,
Melisa Savich
,
Shengtuo Hu
,
Zhuoqing Morley Mao
· 0 min read
Abstract
In Autonomous Vehicle (AV) platooning, vehicles queue up with minimal following distances for improved traffic density and fuel economy. If one vehicle is compromised and suddenly brakes, these AVs will most likely be unable to prevent a collision. In this work, we propose a proactive approach to platooning security: Autonomous Vehicle contracts, in which AVs are architected to use secure enclaves to enforce agreed-upon driving rules, such as a restriction not to brake harder than a certain threshold while the contract is in effect. We explore whether AV contracts will be feasible in worst-case emergency situations while simultaneously under attack, when it is imperative to return full autonomy to AVs as soon as possible. Through our prototype contract implementation using Intel SGX enclaves, including measurement from real-world testing of wireless On-Board Units (OBUs), we show that AV contracts can be quickly and safely terminated in the event of an emergency while retaining a false positive rate of under 0.001% per 10 hours of use. We find that individual autonomy can be returned to the vehicles of an 8-vehicle platoon under contract within 1.5 seconds of an attack, including both detection and safe vehicle separation. Smaller platoons are even quicker. Consequently, automobile manufacturers may find the additional safety offered by AV contracts to provide a net benefit.
Type
Publication
In 2018 IEEE Vehicular Networking Conference
Shibo Chen
Authors
PhD Candidate
My research interests include heterogeneous computation, datacenter architecture and agile hardware design. My advisor is Prof. Todd Austin. I am a student in ADA (The Center for Applications Driving Architectures) and CELAB (Computer Engineering Lab) at University of Michigan.