Securing Cloud-Assisted Connected and Autonomous Vehicles: An In-Depth Threat Analysis and Risk Assessment

Al Tariq Sheik, Carsten Maple, Gregory Epiphaniou, Mehrdad Dianati

TL;DR

This paper tackles the critical security challenges of Cloud-Assisted Connected and Autonomous Vehicles (CCAVs) by providing a comprehensive threat analysis and risk assessment. It systematically identifies vulnerabilities, quantifies risks, and proposes a novel defense taxonomy, offering a much-needed framework for securing these complex, interconnected systems.

What Problem Does This Paper Address?

The increasing integration of cloud services, edge computing, and autonomous driving technologies in vehicles (CCAVs) creates a vast and complex attack surface. Existing threat analysis and risk assessment (TARA) methodologies often fall short in capturing the unique challenges of CCAVs, such as intricate hardware-software interactions, rapid technological advancements, and heterogeneous standards. This leads to poorly defined threat boundaries and reduced TARA efficacy, leaving these safety-critical systems vulnerable to sophisticated cyberattacks. The paper aims to bridge this gap by providing a systematic and in-depth analysis tailored to the CCAV ecosystem.

Methods

The authors employ a multi-faceted approach to threat analysis and risk assessment. They begin by systematically evaluating existing TARA methods and select the STRIDE (Spoofing, Tampering, Repudiation, Information Disclosure, Denial of Service, Elevation of Privilege) threat model for identifying threats and the DREAD (Damage, Reproducibility, Exploitability, Affected Users, Discoverability) methodology for quantifying risks.

Key methodological steps include:

  • System Architecture Analysis: Deconstructing CCAV systems into their constituent tiers (vehicle, edge cloud, core cloud) and identifying trust domains.
  • Data Flow Diagrams (DFDs): Creating DFDs to visualize data flow and identify potential entry/exit points for threats.
  • STRIDE/DREAD Analysis: Applying STRIDE to categorize threats and DREAD to assess their severity, resulting in a risk score.
  • Attack Tree Construction: Developing attack trees to visually represent potential attack pathways and their impact on CCAV systems.
  • Defense Taxonomy Development: Proposing a novel defense taxonomy that maps identified attack mechanisms to hardware and software components, guiding the development of countermeasures.

Datasets

Not discussed in the paper. The authors mention using data collected from literature and real-life incidents from previous research, specifically referencing [8] for a survey of 63 real-life attacks between 2015 and 2022.

Key Insights and Ideas

  • Complexity Amplifies Risk: The interconnected nature of CCAVs, while enabling advanced functionalities, significantly expands the attack surface, making them inherently more vulnerable than traditional vehicles.
  • STRIDE/DREAD is a Foundation, Not a Panacea: While STRIDE and DREAD are valuable tools, the paper highlights their limitations in fully capturing the nuances of CCAV security, particularly regarding complex adversarial behaviors and hardware-software interactions.
  • Trust Domains are Interconnected: A single threat can impact multiple trust domains within a CCAV system, emphasizing the need for a holistic security approach.
  • "Warning" Risks Dominate: The study reveals a significant proportion of "Warning"-level risks (medium risk) in both CCAV and edge/cloud systems, indicating ongoing security challenges that require layered strategies.
  • Privacy is Not an Afterthought: The paper underscores the growing importance of privacy as a security requirement, noting its increased compromise in edge/cloud systems and the need for privacy-by-design approaches.
  • Hardware-Software Interplay is Crucial: The security of CCAVs hinges on securing both hardware and software components, and effective countermeasures must address this interplay.

Key Results

  • The study identified a significant number of threats, categorizing them by risk level (Low, Medium, High) for CCAVs and edge/cloud systems.
  • High-risk threats were found to predominantly affect trust domains like Vehicle Sensors (V-TD4), Physical Input/Output (V-TD5), Energy System (V-TD8), and Data Analysis (V-TD10) in CCAVs.
  • For edge/cloud systems, high-risk threats significantly impacted Microservices (E-TD2, C-TD3), APIs (E-TD3, C-TD4), and Data Storage (E-TD6, C-TD5).
  • Confidentiality, Integrity, and Availability (CIA triad) were identified as the most frequently compromised security requirements, accounting for a substantial portion of security compromises in both CCAV and edge/cloud systems.
  • The paper developed a defense taxonomy that maps identified attack mechanisms to specific trust domains and proposes a set of countermeasures, categorized by hardware and software security.

Limitations and Future Work

The authors acknowledge several limitations:

  • TARA Limitations: Existing TARA methods struggle to precisely represent complex adversarial behavior, especially in targeted or multistage attacks that exploit physical components. They may not fully accommodate the unique security challenges of rapidly evolving CCAV technology.
  • Hardware-Software Interaction: The current analysis techniques often overlook the intricate relationship between hardware and software components, failing to fully capture vulnerabilities arising from their interaction.
  • Defining Conformance: Accurately characterizing how a CCAV system conforms to security requirements is challenging due to the difficulty in defining, verifying, and validating conformance.
  • Future Work Recommendations: The authors suggest future research should aim to develop and employ alternative modeling techniques to analyze security threats in CCAV systems, better capture intricate vulnerabilities, and inform the latest countermeasures. They also propose integrating advanced persistent threats (APTs) and machine learning (ML)/deep learning (DL) for threat ranking and analysis.

Summary updated

BibTeX
@inproceedings{sheik2024securing,
  author = {Al Tariq Sheik and Carsten Maple and Gregory Epiphaniou and Mehrdad Dianati},
  title = {{Securing Cloud-Assisted Connected and Autonomous Vehicles: An In-Depth Threat Analysis and Risk Assessment}},
  booktitle = {Sensors},
  year = {2024},
  url = {https://openreview.net/forum?id=t4mwbhxZba},
}
Figure