An Analytical Framework for Modeling and Synthesizing Malicious Attacks on ACC Vehicles

TL;DR

This work tackles the vulnerability of adaptive cruise control (ACC) vehicles to malicious, covert cyberattacks by developing an analytical framework to model and synthesize false-data injection attacks on ACC sensor measurements within mixed traffic containing HDVs. Using a car-following basis, the authors integrate attacker mappings $g_1$ and $g_2$ that generate additive or multiplicative sensor corruptions, and characterize these attacks under rational driving constraints to ensure stealth and avoid collisions. The framework is instantiated with IDM for HDVs and the OVRV ACC model, with explicit conditions and illustrative examples demonstrating how small, stealthy perturbations can cause substantial disruptions in traffic flow and increase energy consumption, as validated by numerical simulations. The approach provides a physically interpretable set of candidate attacks and a foundation for developing detection and mitigation strategies in future work, including extensions to varying automation levels and V2V/CACC contexts.

Abstract

While emerging adaptive cruise control (ACC) technologies are making their way into more vehicles, they also expose a vulnerability to potential malicious cyberattacks. Previous research has typically focused on constant or stochastic attacks without explicitly addressing their malicious and covert characteristics. As a result, these attacks may inadvertently benefit the compromised vehicles, inconsistent with real-world scenarios. In contrast, we establish an analytical framework to model and synthesize a range of candidate attacks, offering a physical interpretation from the attacker's standpoint. Specifically, we introduce a mathematical framework that describes mixed traffic scenarios, comprising ACC vehicles and human-driven vehicles (HDVs), grounded in car-following dynamics. Within this framework, we synthesize and integrate a class of false data injection attacks into ACC sensor measurements, influencing traffic flow dynamics. As a first-of-its-kind study, this work provides an analytical characterization of attacks, emphasizing their malicious and stealthy attributes while explicitly accounting for vehicle driving behavior, thereby yielding a set of candidate attacks with physical interpretability. To demonstrate the modeling process, we perform a series of numerical simulations to holistically assess the effects of attacks on car-following dynamics, traffic efficiency, and vehicular fuel consumption. The primary findings indicate that strategically synthesized candidate attacks can cause significant disruptions to the traffic flow while altering the driving behavior of ACC vehicles in a subtle fashion to remain stealthy, which is supported by a series of analytical results.

Figures (6)

• Figure 1: A graphic representation of mixed traffic consisting of ACC vehicles and HDVs, where ACC vehicles are vulnerable to cyberattacks while HDVs are not susceptible to such attacks.
• Figure 2: A graphic illustration depicting a 10-vehicle string in mixed traffic, where the second vehicle is an ACC vehicle susceptible to potential attacks.
• Figure 3: Speed profile of all vehicles for Case 1 -- Case 6, with attacks occurring during the period of $I = \left[50, 80\right]$ sec. The exact form of candidate attacks for each case is given in its corresponding figure caption, e.g., $g_1 = 0.1\sin(s) - 0.1s$ and $g_2 = 0.1\sin(\Delta v) - 0.1\Delta v$ for Case 1. It is worth noting that the attacks shown in Case 1 -- Case 3 are taken from the set $\mathcal{C}$, while those of Case 4 -- Case 6 are drawn from the complement of $\mathcal{C}$.
• Figure 4: Displacement profile of all vehicles for Case 1 -- Case 6, with attacks occurring during the period of $I = \left[50, 80\right]$ sec. The simulation settings of each case are same as those of that case shown in Figure \ref{['speed_profile']}.
• Figure 5: ASV of Case 1, Case 2, and Case 3, with attacks launched from the set $\mathcal{C}$. The value of ASV is calculated for vehicles #2 -- #10 over the period of $\left[50, 200\right]$ sec.

Theorems & Definitions (4)

• Remark 3.1
• Remark 3.2
• Remark 3.3
• Remark 3.4