The advent of connected autonomous vehicles provides opportunities for safer, smoother, and smarter transportation. However, broadcasting information to surrounding vehicles and infrastructures risks security and privacy. Moreover, control decisions relying on such information are vulnerable to malicious attacks. In this paper, we propose a cooperative control strategy incorporating with efficient multi-party computation (MPC). In an effort to perform secure MPC without third-party authentication while reducing latency, we integrate a function secret sharing scheme with a distributed oblivious random access memory. We further design an adaptive proportional-derivative controller to increase resilience toward latency and adversaries. Theoretical foundations and limitations are also discussed.
Since the first competition of autonomous vehicles hosted by
the Defense Advanced Research Projects Agency (DARPA)
Grand Challenge in 2005
On the other hand, connected vehicle techniques are also
being deployed to improve the safety and mobility of our
transportation system by enhancing situational awareness
and traffic state estimation through vehicle-to-vehicle (V2V)
and vehicle-to-infrastructure communications, which can
enable applications like cooperative collision warning,
providing traffic signal status information in real time and so
on
More recently, interesting opportunities appear through
the utilization of techniques from connected vehicles to
autonomous vehicles. The connectivity allows an autonomous
to have more detailed knowledge of the environment. The
sensing capability of the autonomous vehicle is hence
further expanded. A platoon formed by connected and
autonomous (CAVs) on the road can increase the capacity,
reduce energy consumption and improve safety. It was
predicted that the transition from the current human-driven
vehicles to a fully CAV traffic environment require a few
decades
In reality, V2V communications are unreliable due to
factors such as interference, network congestion, and malicious
attacks; in the worst case, V2V networks undergo
Byzantine failures
The rest of this paper is organized as follows: in Section 2
we introduce necessary background; Section 3 presents the
AutoMPC model; future works are discussed in Section 4;
we leave the experimental section and more theoretical
results in the full paper.
To formalize the problem, we suppose a multi-agent
system in which each party i has a secret input xi and a
function f (x1; x2; : : : ) can be jointly evaluated. Secure
multiparty computation (MPC) is a mechanism to ensure that each
party known the output of the function f while being
unaware of others’ inputs. Two-party computation (2PC) is a
special case of MPC, which was first introduced by
Oblivious random access memory (ORAM)
Secret sharing
Inspired by existing works
Eval(i; ki; x) is a polynomial-time evaluation algorithm, which on input i 2 [m] (party index), ki (key defining fi : f0; 1gn ! G), and x 2 f0; 1gn (intput for fi) outputs a group element yi 2 G (the value of fi(x), the i-th share of f (x)).
The setting of FSS ensures correctness and security that
each party’s key cannot individually reveal any information
of f
To mitigate the latency trade-offs given by MPCs and increase resilience towards adversaries, we propose an adaptive proportional-derivative (PD) controller based on a twopredecessor-following scheme as shown in Figure 2, in which we assume all CAVs in the platoon to be identical, forming a homogeneous vehicle string. Below is the control command
Ui(s) = Ub;i(s) + Uf;i 1(s) + Uf;i 2(s) (1) which consists of control feedback Ub;i from the error Ei and two extra feedforward terms Uf;i 1 and Uf;i 2 from the acceleration rates Xi 1 and Xi 2, respectively. Xi is the position output, Xi 1 is the feedback position information from the immediate predecessor. Ki is the feedback controller which generates a control command to rectify the error. Gi represents the ideal longitudinal vehicle dynamics. Hi denotes spacing policy (e.g., CD and CTH), and F1;i and F2;i are feedforward filters to process the acceleration information from the corresponding predecessor vehicles. and are indicators for the success of V2V communications ( and are equal to 1 for a successful communication between the CAV and the corresponding predecessor vehicles, and 0 otherwise). These terms will be explained in detail later. 4
The AutoMPC model leverages advances in cryptography to control theory for safer, smoother, and smarter transportation. The contributions lie in several ways: (i) security and privacy are guaranteed via a MPC scheme, without the presence of third-party authentication; (ii) the efficiency of the MPC is achieved by a distributed oblivious RAM and a function secret sharing scheme, and thereby avoids the homomorphic encryption approach which is computationally expensive; and (iii) an adaptive proportional-derivative controller is proposed to increase the resilience toward latency and adversarial attacks. Preliminary experimental results also validate above findings by comparing control performances in speed, spacing, and acceleration rate. Theoretical properties in security and control as well as more experimental results will be discussed in the full paper.
The authors thank Jian Wang, Yuntao Guo, Chaojie Wang, and Anye Zhou for insightful discussions and anonymous reviewers for their helpful comments.
Li, T. 2018. Modeling uncertainty in vehicle trajectory prediction in a mixed connected and autonomous vehicle environment using deep learning and kernel density estimation. In the Fourth Annual Symposium on Transportation Informatics.
Lin, L., and Li, W. 2018. A compressive sensing approach for connected vehicle data capture and recovery and its impact on travel time estimation. arXiv preprint arXiv:1806.10046.
Lin, L.; Wang, Q.; Huang, S.; and Sadek, A. W. 2014. Online prediction of border crossing traffic using an enhanced spinning network method. Transportation Research Part C: Emerging Technologies 43:158–173.
Lin, L.; Wang, Q.; and Sadek, A. W. 2015. A novel variable selection method based on frequent pattern tree for real-time traffic accident risk prediction. Transportation Research Part C: Emerging Technologies 55:444–459.
Lin, L.; Wang, Q.; and Sadek, A. W. 2016. A combined m5p tree and hazard-based duration model for predicting urban freeway traffic accident durations. Accident Analysis & Prevention 91:114–126.
Lin, L. 2018. Efficient collection of connected vehicle data based on compressive sensing. arXiv preprint arXiv:1806.02388.
Lu, S., and Ostrovsky, R. 2013. Distributed oblivious ram for secure two-party computation. In Theory of Cryptography. Springer. 377–396.