Framework implementation

In this section, we describe the implementation of the framework stages for our evaluation including the traffic scenarios we use, the pseudonym strategies we evaluate, and the attacker’s strategy and coverage. The results are presented separately in Section 3.7.

Model mobility

An evaluation like ours can only provide meaningful guidance if it is conducted in realistic traffic scenarios, both with regard to the shape of the road network and the realism of the traffic flows. While some real-world mobility traces are available, they were unsuitable for a large-scale analysis like ours because they only cover a small fraction of the overall traffic (e.g., taxis or buses) or were recorded in confined scenarios (e.g., on corporate grounds).

We use two different, synthetic but realistic, large-scale traffic scenarios. To evaluate pseudonym changes in an urban scenario, we chose the freely-available LuST scenario [49] of 24 hours of synthetic traffic in the city of Luxembourg. In order to focus on urban traffic, we did not include the “external mobility” part of the scenario, i.e., traffic that has both its origin and destination outside the city area and only passes through on the highway. See Figure 3.3 for an overview of the LuST scenario. For evaluation in a highway scenario, we

For evaluation in urban traffic, we use the LuST scenario [49] of 24hours of traffic in the city of Luxembourg

Figure 3.3 For evaluation in urban traffic, we use the LuST scenario [49] of 24hours of traffic in the city of Luxembourg.

created a simulation of 24 hours of synthetic but realistic traffic on a 45 km highway segment near Stuttgart, Germany. We built the road network based on OpenStreetMap data. The information about traffic volumes were extracted from data of traffic counting stations. We used hourly data collected by the German Federal Highway Research Institute [72] and annual average daily traffic data collected by the Road Traffic Center Baden-Wurttemberg [156]. Furthermore, we differentiated between four vehicle types (passenger cars, motorbikes, buses, and trucks). Based on this data, we created traffic flows for 24 hours, divided into workdays and holidays. See Figure 3.4 for an overview of the highway scenario.

For evaluation in a highway scenario, we created a simulation of 24 hours of realistic traffic on a 45 km highway segment near Stuttgart, Germany

Figure 3.4 For evaluation in a highway scenario, we created a simulation of 24 hours of realistic traffic on a 45 km highway segment near Stuttgart, Germany.

Table 3.1 Total number of trips and average trip duration in the urban and highway simulation scenarios at low traffic (1 — 4 a.m.) and high traffic (7 — 10 a.m.).

Scenario

Number of trips

Average trip duration

Urban, low traffic

1226

561 s

Urban, high traffic

56419

755 s

Highway, low traffic

3 545

506 s

Highway, high traffic

34414

526s

Traffic simulations are conducted using SUMO [114], a microscopic traffic simulator, version 0.25. We use the default car-following model, which is an adaption of the model by Kraufi [115] and provides stochastic driving behavior for individual vehicles. For most realistic driving behavior, we set real-life acceleration and deceleration parameters for each of our four vehicle classes. Instead of running the full 24 hour scenarios, we created a low traffic variant (between 1 and 4 a.m.) and a high traffic variant (between 7 and 10 a.m.) for both scenarios. Table 3.1 displays the number of trips and average trip duration for each of our scenarios.

The traces are recorded with a frequency of 1 Hz, which symbolizes the beaconing interval. We reduced the standard frequency of up to 10 Hz to keep the run time of simulations and the size of files within practical limits.

 
Source
< Prev   CONTENTS   Source   Next >