3 March 2026 in IEEE Transactions on Vehicular Technology
The growing Internet of Vehicles (IoV) requires efficient edge data management, where collaborative cache sharing between Roadside Units (RSU) and vehicles is crucial for reducing latency in applications like real-time navigation and collision avoidance. However, direct sharing of sensitive data such as vehicle locations and requests introduces serious privacy risks. To address this, we propose a lightweight privacy-preserving cache sharing scheme with label protection for vehicular networks. We first design an Oblivious Homomorphic XOR-PRF (Ohx-PRF) that employs key pre-distribution and homomorphic aggregation to eliminate oblivious transfer interactions. Building on this, we construct a non-interactive labeled delegated multi-party private set intersection (LDMPSI) protocol, where each data subset carries a classification label. Finally, we introduce a cache sharing scheme based on LDMPSI that uses efficient multi-point OPRF with only symmetric cryptography. By delegating computations to servers and processing data and labels in ciphertext, RSU can identify and store frequently requested content with minimal overhead. Formal analysis proves the proposed protocol and scheme are secure under the semi-honest model. Experiments show our scheme achieves over 85% cache hit rate, while LDMPSI reduces total runtime by 79% and communication overhead by over 80% compared to existing protocols, enhancing resource utilization and response capabilities in vehicular networks without compromising privacy.
30 April 2026 in IEEE Internet of Things Journal
While privacy leakage and inefficient reputation management are known challenges for vehicular crowdsensing, existing solutions suffer from the following limitations: 1) Current location privacy-preserving matching schemes based on generalization or pseudonyms cannot achieve accurate encrypted location overlap determination; 2) Existing reputation verification mechanisms lack zero-knowledge proof support, requiring excessive data exposure during verification and consequently increasing privacy risks; 3) Inadequate defense against collusion attacks, which disrupt reputation values. To address these issues, this paper proposes a lightweight privacy-preserving scheme with efficient reputation management (LPRM) for vehicular crowd-sensing: 1) A lightweight privacy-preserving location matching (LPLM) algorithm that achieves location privacy protection and completes location matching through encryption and ciphertext overlap detection; 2) A lightweight privacy-preserving reputation verification (LPRV) algorithm that employs Paillier homomorphic encryption to protect reputation values and enables leak-free legal and equivalence verification of reputation through zero-knowledge proofs; 3) A five-factor dynamic reputation update (FFRU) model comprising a collusion detection mechanism and a dynamic penalty mechanism. The collusion detection mechanism identifies coordinated user behavior through feedback similarity analysis, while the dynamic penalty mechanism adaptively adjusts the penalty factor based on collusion attack level, using a polynomial penalty function to rigorously penalize malicious users without over-penalizing minor infractions. Through theoretical analysis, we have demonstrated the advantages of the LPRM scheme in terms of property, privacy, and security. The performance evaluation proves that the LPRM scheme can decrease the computation and communication overheads by 30% and 71% respectively compared to the existing optimal scheme, and enhance the effectiveness and robustness of the scheme.
1 July 2026 in IEEE Transactions on Computers