Virtual Pseudonym-Changing and Dynamic Grouping Policy for Privacy Preservation in VANETs

Ikram Ullah, Munam Ali Shah, Abid Khan, Carsten Maple, Abdul Waheed

TL;DR

This paper introduces a novel Dynamic Grouping and Virtual Pseudonym-Changing (DGVP) scheme to enhance location privacy in Vehicular Ad-hoc Networks (VANETs). By dynamically grouping vehicles based on similar status and cooperatively changing pseudonyms, DGVP significantly improves anonymity and reduces traceability, even in challenging high-speed and low-traffic scenarios.

What Problem Does This Paper Address?

Vehicular Ad-hoc Networks (VANETs) are crucial for Intelligent Transportation Systems (ITS), enabling vehicles to share information about road status, safety, and traffic. However, this communication, often through beacon messages containing vehicle IDs, positions, and speeds, creates a significant privacy risk. Adversaries can exploit this data to track vehicles, infer driver behavior, and potentially cause harm. Existing solutions like pseudonym-changing and mix zones have limitations, particularly in high-speed or low-traffic conditions, and can negatively impact VANET application functionality. This paper addresses the need for a robust, dynamic, and efficient privacy-preserving scheme that works effectively across diverse traffic conditions without compromising essential VANET services.

Methods

The proposed Dynamic Grouping and Virtual Pseudonym-Changing (DGVP) scheme has several key components:

  • Dynamic Grouping: Vehicles with similar status (velocity range, direction) are dynamically grouped. A Group Head (GH) is selected to manage the group and coordinate pseudonym changes. This grouping is based on road context information.
  • Cooperative Pseudonym Changing: Within a group, vehicles cooperatively change their pseudonyms. This is triggered when a certain "neighbor threshold" is met, ensuring a higher level of anonymity.
  • Virtual Pseudonym Update Process: In low-traffic scenarios where the neighbor threshold might not be met, a "virtual pseudonym update" is employed. This involves vehicles generating randomized versions of pseudonyms to create uncertainty for adversaries.
  • Position-Mixing with LBS: When vehicles interact with Location-Based Services (LBS), a position-mixing method is used. Vehicles exchange their position coordinates with a randomly selected neighbor within the group, effectively blending their precise location with that of their neighbor to confuse adversaries.
  • Formal Modeling: The scheme is formally modeled and specified using High-Level Petri Nets (HLPN) to analyze its behavior and properties.

Datasets

Not discussed in the paper. The authors use a simulation environment (SUMO and NS-2) with generated traffic data.

Key Insights and Ideas

  • Cooperative Privacy is Key: The paper strongly emphasizes that cooperative pseudonym changing among a group of vehicles is far more effective than individual efforts for robust location privacy.
  • Adaptability is Crucial: The DGVP scheme's ability to adapt to different traffic densities (using cooperative changes in high density and virtual changes in low density) is a significant advantage over static methods.
  • Virtual Pseudonyms for Sparse Scenarios: The introduction of a "virtual pseudonym update" process is a clever way to maintain privacy even when there aren't enough nearby vehicles for traditional cooperative changes.
  • Position Mixing Thwarts LBS Tracking: The position-mixing technique effectively obfuscates a vehicle's true location when interacting with LBS, a common point of vulnerability.
  • Balancing Privacy and Functionality: A core insight is that privacy mechanisms can be implemented without significantly distorting critical VANET data, thus preserving application quality.

Key Results

The paper presents simulation results comparing DGVP with existing schemes (CPS and CLPS).

  • Anonymity: DGVP significantly improves the Anonymity Set Size (ASS) compared to CPS and CLPS, showing higher anonymization across various time periods and traffic densities. For instance, at 300 seconds, DGVP achieves an ASS of around 70, while CPS and CLPS are significantly lower.
  • Entropy: DGVP demonstrates higher entropy, indicating greater adversary confusion in identifying target vehicles. This is attributed to efficient group management and cooperative pseudonym changes.
  • Traceability: The proposed scheme achieves lower location traceability, meaning it's harder for adversaries to track vehicles. At 300 seconds, DGVP's traceability is around 0.4, while CPS and CLPS are higher.
  • Computation Cost: DGVP shows a modest increase in computation latency compared to CLPS but is lower than CPS. This is a trade-off for enhanced privacy.
  • Successful Pseudonym Changes: DGVP shows a higher rate of successful pseudonym changes over time, especially compared to CPS and CLPS, due to its cooperative approach.

Limitations and Future Work

  • Adversary Model: While the paper considers a robust adversary, it primarily focuses on external passive adversaries. The impact of more sophisticated or active adversaries might warrant further investigation.
  • Scalability: While the scheme is designed for dynamic environments, extensive testing on extremely large-scale networks with highly dynamic group memberships could be beneficial.
  • Road Network Diversity: The authors mention planning to conduct more experiments on diverse road networks to explore other parameters for vehicle location privacy.
  • Impact on Specific VANET Applications: While the paper claims minimal impact on VANET applications, a deeper analysis of how DGVP affects specific safety or infotainment features in real-world scenarios could be valuable.

Summary updated

BibTeX
@inproceedings{ullah2021virtual,
  author = {Ikram Ullah and Munam Ali Shah and Abid Khan and Carsten Maple and Abdul Waheed},
  title = {{Virtual Pseudonym-Changing and Dynamic Grouping Policy for Privacy Preservation in VANETs}},
  booktitle = {Sensors},
  year = {2021},
  url = {https://openreview.net/forum?id=XLsU31SXL2},
}
Figure