Smart healthcare refers to the innovative medical model that enhances the efficiency and quality of medical services through the use of modern information technology. While smart healthcare brings convenience to both patients and doctors, it also raises issues of medical data security and privacy protection. In smart healthcare, secure identity authentication not only requires the verification of user identities but also ensures that user identity information is not leaked. At the same time, it requires an efficient authentication process to cope with resource-constrained situations. Existing schemes have addressed the issues of identity authentication and the confidentiality of identity information, but they have significant overhead and are not suitable for resource-constrained situations. Additionally, the schemes themselves are vulnerable to attacks. To address these issues, we propose a multi-party authentication scheme for smart healthcare. First, based on certificateless public key cryptography(CLPKC), a certificateless signature scheme(CLS) is introduced, which eliminates the need for complex bilinear pairings and map-to-point hash function computations, enabling batch authentication for multiple users. Second, the scheme incorporates the user’s contextual environment, ensuring secure user identity authentication only under safe access conditions. Finally, blockchain technology is utilized to record access logs, enabling user traceability. Security analysis proves that the proposed scheme can meet security requirements such as conditional privacy protection, unlinkability, resistance to replay attacks, and resistance to collusion attacks. Compared with existing schemes, this scheme has lower computational overhead and communication costs, making it more suitable for lightweight multi-party authentication in smart healthcare scenarios.

With the development of the Internet of Vehicles (IoV), the rapid growth of smart vehicles has generated vast amounts of data, which are of significant value for training intelligent IoV application models. Traditional methods of intelligent model training require centralized collection of raw data, consuming significant communication resources and posing issues such as privacy leakage. Federated learning offers an approach that allows multiple participants to collaboratively train models while protecting data privacy. However, in practical application scenarios, the data among vehicles often exhibits the characteristics of non-independent and identically distributed (Non-IID), which can lead to significant performance differences of the global model across different vehicles, thus posing a challenge to achieving fair collaboration among vehicles. Existing solutions tend to pursue the overall performance of all users, while neglecting the issue of fairness between individual users, which may result in significant differences in accuracy between different users. To address this issue, this paper proposes a fair federated learning algorithm that combines cross-entropy and marginal loss to assess the influence of the vehicle’s local model against the global model, and dynamically adjust the aggregation weight and achieve fair federated learning accordingly. This method was compared with existing methods on multiple datasets, and the experimental results show that the proposed method is superior to other methods in terms of fairness.

In the context of vehicular crowdsensing within the Internet of Vehicles (IoV), data aggregation techniques enable the computation and analysis of sensing data to extract valuable insights and improve transmission efficiency. However, sensing data and aggregation results often contain sensitive information about terminal vehicles, posing risks of privacy leakage. Existing privacy-preserving data aggregation schemes typically employ homomorphic encryption or bilinear pairing operations to ensure both privacy protection and integrity verification, which incur significant computational overhead. Moreover, when aggregation nodes are untrusted, it becomes challenging to verify the correctness of the aggregated results. To address these challenges, this paper proposes a lightweight and verifiable multi-dimensional data aggregation privacy-preserving scheme for vehicular crowdsensing. In the data generation phase, a blinding factor is introduced to obfuscate the sensing data, and secret sharing is employed to split the obfuscated data into multiple shares. This approach ensures data privacy and resists collusion attacks among internal vehicles. A lightweight signature aggregation method is integrated to verify multiple digital signatures in a single computation, significantly reducing the computational and communication costs associated with integrity verification. In the data recovery phase, the original data can be restored by removing the blinding factor, eliminating the need for traditional decryption algorithms and thereby reducing computational overhead. In the aggregation result verification phase, a homomorphic commitment mechanism is adopted. Users generate commitments of the obfuscated sensing data and upload them to the blockchain, effectively addressing the issue of untrusted roadside units and traffic management centers in real-world applications and ensuring the correctness of the aggregated results. Experimental results demonstrate that the proposed scheme reduces computational overhead by more than 50% compared to existing methods, exhibiting higher efficiency and better adaptability.

With the popularity of social networks, a great variety of new social applications have been generated for impromptu group formation and communications. Among those applications, the subgraph matching has become a hot research area in social networks. Due to the huge cost of managing and computing graph data, it may have to outsource the computations to the cloud server. However, the most critical problem is that the cloud server leaks the graph information during the processing of the graph data, and the external attackers modify the graph information during the transmission on the public channel. Thus, confidentiality and authentication have been critical attributes in the subgraph matching query service. In this article, we present an efficient and privacy-preserving subgraph matching scheme with authentication in social networks. Using the proposed scheme, the cloud can accomplish the subgraph matching query process without obtaining any sensitive information about the users. Additionally, we achieve data integrity verification and user authentication. Each receiver can verify if the received messages come from the legal sender and have not been tampered. The detailed security and efficiency analysis show that the proposed scheme not only satisfies security requirements but also achieves high-efficiency in local users, and it is suitable for many practical applications.

Due to wealthy structure and semantic information expressed by a graph, the graph is frequently employed in numerous social applications to show social relationships. Among those important applications, the private graph intersection operation plays an important part in social networks. Because of the high cost of managing graph data and the computational difficulty of graph intersection operation, delegating the computations to the cloud server (CS) is an attractive alternative. However, when the CS is untrusted or compromised by some adversaries, the results that the CS returns can not be guaranteed to be correct. In such cases, it may have serious consequences for the application functionality. In this article, we present an efficient and privacy-preserving verifiable graph intersection scheme with cryptographic accumulators in social networks. Using the proposed scheme, we construct the framework to provide secure verifiable graph intersection operation in an untrusted cloud, and the requester can verify the correctness of the graph intersection result that the CS returns. Additionally, the data owners' graph data privacy and user authentication are well protected. The detailed correctness proof and performance analysis show that the proposed scheme is secure and feasible. Thus, our scheme is appropriate for many practical applications.