Game-Theoretic Framework for Private Data Sharing in Vehicular Networks

A Stackelberg-based hybrid architecture for privacy-preserving vehicular data markets

Yousef AlSaqabi; Yinan Zhou; Faisal Nawab; Bhaskar Krishnamachari· 2026· DOI 10.48550/arXiv.2606.22115

The core problem

Decentralized vehicular data collection systems face a fundamental tension: while aggregated sensor data can enable valuable services, individual vehicles risk privacy exposure. The authors address this by introducing a hybrid architecture that separates data supply, processing, and consumption. The framework comprises four key entities: vehicles that generate sensor data, independent servers that perform secure multiparty computation (SMPC), a coordinator node that orchestrates data flow, and data consumers that provide economic incentives. The central privacy guarantee is that only the data consumer can access the fully aggregated data, never individual raw data. This design significantly reduces privacy risks while enabling a functional data market. The work integrates Stackelberg competition from game theory to model the strategic interactions between vehicles (followers) and the data consumer (leader), allowing vehicles to make participation decisions based on perceived privacy risks and offered incentives.

The authors evaluate the framework using real-world vehicular location data. They quantify privacy risks by measuring the accuracy of path reconstruction attacks when an adversary has access to a subset of shared data. Key findings include: - The SMPC-based aggregation effectively prevents individual data exposure, as the adversary cannot reconstruct paths from the aggregated output alone. - The Stackelberg incentive mechanism successfully incentivizes participation while maintaining privacy: as privacy risk increases, vehicles require higher incentives, and the consumer adjusts payments accordingly. - Simulations show that the framework supports an active data market, with a stable equilibrium where a sufficient number of vehicles participate. - The privacy-utility trade-off is tunable via the incentive structure: higher payments lead to more data sharing but also higher privacy costs for vehicles. The experiments demonstrate that the proposed system is robust in preserving privacy while enabling profitable data sharing. The authors also explore the dynamics of vehicle participation under varying privacy sensitivities and incentive levels.

Decentralized vehicular data collection systems face a fundamental tension: while aggregated sensor data can enable valuable services, individual vehicles risk privacy exposure. The authors address this by introducing a hybrid architecture that separates data supply, processing, and consumption. The framework comprises four key entities: vehicles that generate sensor data, independent servers that perform secure multiparty computation (SMPC), a coordinator node that orchestrates data flow, and data consumers that provide economic incentives. The central privacy guarantee is that only the data consumer can access the fully aggregated data, never individual raw data. This design significantly reduces privacy risks while enabling a functional data market. The work integrates Stackelberg competition from game theory to model the strategic interactions between vehicles (followers) and the data consumer (leader), allowing vehicles to make participation decisions based on perceived privacy risks and offered incentives.

The proposed framework is built on a hybrid architecture that combines decentralized data collection with secure computation. The key components are:

The paper's main contribution is the integration of game theory with privacy-preserving computation to create a practical marketplace for vehicular data. The Stackelberg model captures the asymmetric roles of the data consumer and vehicles, providing a realistic representation of incentives. The use of SMPC ensures that privacy is not merely a policy but a cryptographic guarantee. However, the framework assumes that vehicles can accurately assess their privacy risk, which may not hold in practice. Additionally, the coordinator node is a potential single point of failure or trust; future work could decentralize this role. The authors also note that the value function V (\cdot) is assumed to be known, but in real markets it may be uncertain. Despite these limitations, the empirical validation with real-world data strengthens the practical relevance. The framework aligns with broader trends in privacy-enhancing technologies and decentralized data markets, offering a blueprint for similar systems in other domains such as IoT and smart cities.

CS practitioners and researchers

Opening member content…