Introduction
As the automotive industry transitions toward Level 4 and 5 autonomy, vehicles are evolving from mechanical transport into hyper-connected "mobile data centers." This shift grants vehicles absolute control over driving functions but simultaneously expands their attack surface through Vehicle-to-Vehicle (V2V), Vehicle-to-Infrastructure (V2I), and Vehicle-to-Everything (V2X) communications. Without a structured understanding of these threats, developing robust security architectures is impossible. This study applies the ETSI Threat, Vulnerability and Risk Analysis (TVRA) methodology to map the cybersecurity landscape of fully autonomous vehicles (AVs), identifying critical vulnerabilities that could lead to life-threatening outcomes.
Methodology: The TVRA Framework
The researchers utilized a subset of the ETSI TVRA methodology to systematically evaluate risks. The process began by defining the Target of Evaluation (ToE), which includes the AV itself, its passengers, and the surrounding roadside infrastructure. Assets were categorized into physical (sensors, Roadside Units), logical (data, protocols), and human (passengers) components.
A core part of this analysis is the Threat Reference Model, which identifies specific access interfaces where an adversary might gain entry:
- Interface A: Interaction between the passenger and the vehicle.
- Interface B: Communication between the vehicle and Roadside Units (RSUs).
- Interface E: The connection between RSUs and the broader Internet Domain.
Risk was then quantified using the formula $Risk = Likelihood \times Impact$, where likelihood is determined by the "attack potential" (required time and expertise) and impact focuses on the severity of consequences.
Key Insights: The Human Element and Critical Entry Points
The analysis revealed that the most significant vulnerabilities are often human-centric. The "Human Trojan Horse" concept highlights how a passenger’s compromised mobile device (Interface A) can serve as a bridge to the vehicle's internal systems. Similarly, Interface B (RSUs) is a high-priority entry point for malicious actors. Securing these specific interfaces would yield disproportionate improvements in overall AV security.
Furthermore, "Illusion attacks"—where sensors are manipulated to deliver fake messages—emerged as a top-tier threat. These attacks allow an adversary to deceive neighboring vehicles, potentially causing physical accidents by creating a false perception of road conditions.
Risk Classification and Results
The study mapped various threats onto a risk matrix based on their likelihood and impact.
- Critical Risks: These include Illusion attacks, Sybil attacks (using multiple identities to consume bandwidth), Jamming, Message Injection, and Sensor Spoofing. These require immediate development of countermeasures.
- Major Risks: Threats such as Eavesdropping, Flooding/Spamming, and Masquerading were classified as major. While high-impact, they may have lower likelihood or slightly less fatal immediate consequences than critical risks.
- Human Impact: The study explicitly found that all threats involving human interaction result in "serious consequences," underscoring that social engineering is as vital a security concern as technical exploits.
Conclusion and Future Directions
The primary defense against high-impact data threats is the implementation of robust encryption mechanisms and pseudonymity to protect user privacy. However, the study acknowledges limitations, such as the assumption that the internet backbone is inherently secure and the static nature of the current risk assessment.
Future work must transition from this theoretical risk map to the design and testing of specific countermeasures, such as position verification protocols and hardware-based audit trails, within real-world AV testbeds. As hacking tools become more automated, the likelihood of complex attacks will only increase, making continuous risk evaluation essential.
Summary updated