Resource-Aware Stealthy Attacks in Vehicle Platoons

Basic Information

• Paper ID: 2510.14119
• Title: Resource-Aware Stealthy Attacks in Vehicle Platoons
• Authors: Ali Eslami (Concordia University), Mohammad Pirani (University of Ottawa)
• Classification: eess.SY cs.SY (Systems and Control)
• Publication Date: October 15, 2025 (arXiv submission)
• Paper Link: https://arxiv.org/abs/2510.14119v1

Abstract

Connected and autonomous vehicles (CAVs) are transforming modern transportation through collaborative applications such as vehicle platooning, where multiple vehicles travel in tight formation to improve efficiency and safety. However, the critical dependence on inter-vehicle communication makes platoons highly vulnerable to attacks, where even subtle manipulations can escalate into severe physical consequences. While existing research primarily focuses on defensive mechanisms, there is limited attention to stealthy adversaries aimed at covertly manipulating platoon behavior. This paper provides a new perspective on attack design by demonstrating how attackers can steer platoons toward desired trajectories while remaining undetected. The research outlines the conditions under which such attacks are feasible, analyzes their dependence on communication topology and control protocols, and investigates the resources required by attackers.

Research Background and Motivation

Problem Background

1. Vulnerability of CAV Systems: Vehicle platoons achieve coordinated control through V2X communication, but this connectivity simultaneously introduces cybersecurity risks
2. Severity of Attack Consequences: Due to safety-security coupling characteristics, even subtle attacks can lead to physical accidents
3. Research Gap: Existing research primarily focuses on defensive mechanisms, lacking systematic analysis from the attacker's perspective

Research Motivation

• Vulnerability Assessment Need: Understanding attacker resource requirements helps identify system weaknesses
• Defense Strategy Design: Analyzing attack feasibility provides guidance for designing resilient countermeasures
• Novel Attack Patterns: Paradigm shift from disruptive attacks to manipulative attacks

Limitations of Existing Approaches

• Primarily focus on system destabilization rather than trajectory manipulation
• Lack systematic analysis of attack resource requirements
• Detection mechanisms prone to false positives during transient phases

Core Contributions

1. Novel Stealthy Attack Strategy: Proposes a covert attack framework for leader-follower vehicle platoons where attackers manipulate the platoon to track their own state rather than merely disrupting the formation
2. Multi-Scenario Attack Design: Designs attack signals for both compromised leader communication channels and secure channels, with rigorous analysis of stealth conditions
3. Control Protocol Impact Analysis: Investigates how agent control protocols affect attacker degradation capabilities, demonstrating how control protocols constrain the feasible set of stealthy attacks
4. Resource Requirement Characterization: Systematically analyzes attacker resource requirements, providing guidance for designing resilient vehicle platoons

Methodology Details

Task Definition

Attack Objectives (Definition 3):

1. Maintain stealth: Satisfy residual signal $\lim_{t→∞} r_i^x = 0$
2. Manipulate tracking: Make attacked vehicle subset track attacker state $x_a$ while maintaining desired spacing

Stealth Definition (Definition 2): An attack remains stealthy when, in the presence of attack signal $a_{ji} ≠ 0$, the residual signal satisfies: $\lim_{t→∞} r_i^x = \lim_{t→∞} \sum_{j∈N_i} \|x_i - x_j^c + d_{ij}\| = 0$

System Model

Vehicle Dynamics

Employs a simplified model after feedback linearization: $\tau \dot{a}_i(t) + a_i(t) = u_i(t)$

State-space representation: $\dot{x}_i = Ax_i + Bu_i, \quad i ∈ \{0,...,N\}$

Where: $A = \begin{bmatrix} 0 & 1 & 0 \\ 0 & 0 & 1 \\ 0 & 0 & -\frac{1}{\tau} \end{bmatrix}, \quad B = \begin{bmatrix} 0 \\ 0 \\ \frac{1}{\tau} \end{bmatrix}$

Communication Topology

Employs k-nearest neighbor leader-tracking topology, modeled through directed graph $G = (V, E, A)$, with Laplacian matrix: $L = \begin{bmatrix} 0 & 0_{1×N} \\ L_2 & L_1 \end{bmatrix}$

Controller Design

Dynamic Controller

$u_i = e_i K \sum_{j∈N_i} (x_i - x_j + d_{ij}) + e_i \text{sgn}(K \sum_{j∈N_i} x_i - x_j + d_{ij})$

Where the coupling gain $e_i$ dynamics are: $\dot{e}_i = τ[\sum_{j∈N_i} x_i - x_j + d_{ij}]^T Γ [\sum_{j∈N_i} x_i - x_j + d_{ij}] + τ\|K \sum_{j∈N_i} x_i - x_j + d_{ij}\|_1$

Static Controller

$u_i = c_1 K \sum_{j∈N_i} (x_i - x_j^c + d_{ij}) + c_2 \text{sgn}(K \sum_{j∈N_i} (x_i - x_j^c + d_{ij}))$

Attack Strategy Design

Scenario 1: All Communication Channels Compromised

Attack Signal (Theorem 1): $a_{0i} = x_a - x_0$

Theorem 1: Under Assumptions 1 and 2, attack signal (15) maintains stealth while causing follower vehicles to track the attacker state with desired spacing.

Scenario 2: Leader Communication Channel Secure

Attack on Non-Directly Connected Vehicles (Theorem 2):

• Communication channels from $V_d$ to $V_{nd}$: $a_{ij} = x_a - x_0 + d_i$
• Communication channels from $V_{nd}$ to $V_d$: $a_{ji} = x_0 - d_j - x_a + d_j^c$

Partial Vehicle Attack (Theorem 3): For subset $V_{nd}^a ⊆ V_{nd}$, designs more complex attack signal combinations.

Experimental Setup

Simulation Parameters

• Platoon Size: 1 leader + 4 followers
• Topology: k=2 nearest-neighbor topology
• Time Constant: $τ_i = 0.4$
• Desired Spacing: 20 meters
• Control Gain: $K = [-0.7, -1.2, -0.05]$
• Leader Input: $u_0 = 0.4(0.005\sin(0.1t) + 0.0075\cos(0.1t))$

Test Scenarios

1. Leader communication channel attacked (all followers affected)
2. Leader channel secure, attacking non-directly connected vehicles
3. Selective attack on single follower
4. Attack under static controller

Experimental Results

Main Results

Scenario 1: Comprehensive Attack

• Trajectory Manipulation: All follower vehicles successfully redirect to track attacker trajectory
• Stealth: Residual signals converge to zero, attack remains undetected
• Spacing Maintenance: Vehicles maintain desired 20-meter spacing

Scenario 2: Partial Attack

• Selective Control: Non-directly connected vehicles (vehicles 3 and 4) track attacker
• Directly Connected Vehicles: Continue tracking legitimate leader
• Detection Evasion: Detection possible only in directly connected vehicles at steady state

Scenario 3: Single Vehicle Attack

• Precise Control: Only vehicle 4 tracks attacker
• Other Vehicles: Maintain tracking of leader
• Resource Efficiency: Minimizes attack resource requirements

Scenario 4: Static Controller

• Conditional Attack: Maintains stealth when $γ_2 ≤ γ$
• Detection Tradeoff: Achieves manipulation but becomes detectable when $γ_2 > γ$

Resource Requirement Analysis

According to Table I summary, resource requirements for different attack scenarios:

Related Work

Resilient Control Research

1. Design Resilient Methods: MSR methods require graphs with sufficient r-robustness
2. Control Input Correction: Estimate attack signals for controller design
3. Attack Recovery Methods: Restore system performance after attack detection

Attack Design Research

• Undetectable Attacks: Analyze how to prevent consensus while maintaining stealth
• FDI Attacks: Stealth definitions based on probability distributions
• Graph Condition Studies: Conditions for stealthy FDI attacks causing unbounded consensus errors

Research Gaps

Existing attack research primarily focuses on MAS destabilization, overlooking more subtle and dangerous trajectory manipulation attack categories.

Conclusions and Discussion

Main Conclusions

1. Attack Feasibility: Demonstrates that attackers can achieve trajectory manipulation objectives across multiple scenarios
2. Resource Dependency: Attack success heavily depends on available attacker resources
3. Control Protocol Impact: Different control protocols significantly affect attack feasibility
4. Detection Mechanism Defects: Existing detection mechanisms have blind spots during transient phases

Limitations

1. Model Simplification: Employs linearized vehicle model; actual nonlinear dynamics are more complex
2. Topology Constraints: Primarily targets k-nearest neighbor topology; applicability to other topologies requires verification
3. Detection Mechanisms: Residual generation mechanisms are relatively simple; more sophisticated detection methods may be effective
4. Practical Deployment: Experiments based solely on simulation; lacks real-world environment validation

Future Directions

1. Heterogeneous Platoons: Extend analysis to heterogeneous vehicle platoons
2. Time-Varying Topology: Investigate effects of dynamic topology changes
3. Improved Detection: Develop detection mechanisms effective during transient phases
4. Privacy Protection: Create protocols integrating privacy protection and attack resilience

In-Depth Evaluation

Strengths

1. Problem Innovation: Paradigm shift from disruptive to manipulative attacks has significant importance
2. Theoretical Rigor: Provides complete mathematical proofs and stability analysis
3. Practical Value: Resource requirement analysis provides concrete guidance for defense strategy design
4. Systematic Research: Comprehensive analysis covering multiple attack scenarios and control protocols

Weaknesses

1. Model Limitations: Linearization assumptions may not apply to complex driving scenarios
2. Detection Mechanisms: Residual design is relatively simple, potentially vulnerable to more advanced detection methods
3. Experimental Validation: Lacks verification experiments on real vehicle platforms
4. Countermeasure Research: Insufficient discussion of specific implementation details for defensive mechanisms

Impact

1. Academic Contribution: Opens new research directions for CAV security
2. Practical Value: Provides important reference for autonomous driving system safety design
3. Policy Impact: Can provide technical support for CAV safety standard development
4. Industrial Application: Facilitates development of vehicle-to-everything network security products

Applicable Scenarios

1. Highway Platooning: Applicable to vehicle platoons in structured road environments
2. Logistics Transportation: Coordinated control systems for freight convoys
3. Public Transportation: Safety assessment for intelligent transit systems
4. Military Applications: Safety protection design for unmanned vehicle formations

References

The paper cites 63 related references, primarily including:

1. CAV Security Surveys: Sun et al. (2021) on CAV network security
2. Formation Control: Zheng et al. (2015) on vehicle platoon stability and scalability
3. Resilient Control: LeBlanc et al. (2013) on resilient asymptotic consensus in robust networks
4. Attack Detection: Barboni et al. (2020) on detecting stealthy cyber attacks in interconnected systems

Overall Assessment: This is a high-quality paper with important contributions to CAV security, proposing novel attack perspectives and systematic analytical frameworks that provide valuable insights for both theoretical research and practical applications in the field. Despite certain limitations, its innovation and practical value make it an important reference in the field.