Federated learning (FL) trains a model over multiple datasets by collecting the local models rather than raw data, which can help facilitate distributed data analysis in many real-world applications. Since the model parameters can leak information about the training datasets, it is necessary to preserve the privacy of the FL participants’ local models. Furthermore, FL is vulnerable to poisoning attacks which can significantly decrease the model utility. To settle the above issues, we propose a privacy-preserving and Byzantine-robust FL scheme$\Pi _{\text{P2Brofl}}$that maintains robustness in the presence of poisoning attacks and preserves the privacy of local models simultaneously. Specifically,$\Pi _{\text{P2Brofl}}$leverages three-party computation (3 PC) to securely achieve a Byzantine-robust aggregation method. To improve the efficiency of privacy-preserving local model selection and aggregation, we propose a maliciously secure top-$k$protocol$\Pi _{\text{top}-k}$that has low communication overhead. Moreover, we present an efficient maliciously secure shuffling protocol$\Pi _{\text{shuffle}}$since secure shuffling is necessary for our secure top-$k$protocol. The security proof of the scheme is given and experiments on real-world datasets are conducted in this paper. When the proportion of Byzantine participants is 50%, the error rate of the model only increases by 1.05% while it increases by 23.78% without using our protection.

In mobile crowdsensing (MCS), truth discovery (TD) plays an important role in sensing task completion. Most of the existing studies focus on the privacy preservation of mobile users, and the reliability of mobile users is evaluated by their weights which are calculated based on the submitted sensing data. However, if mobile users are unreliable, the submitted sensing data and their weights are also unreliable, which may influence the accuracy of the ground truths of sensing tasks. Therefore, this article proposes a privacy-preserving and reputation-based truth discovery framework named PRTD which can generate the ground truths of sensing tasks with high accuracy while preserving privacy. Specifically, we first preserve sensing data privacy, weight privacy, and reputation value privacy by utilizing the Paillier algorithm and Pedersen commitment. Then, to verify whether the reputation values of mobile users are tampered with and select mobile users that satisfy the corresponding reputation requirements, we design a privacy-preserving reputation verification algorithm based on reputation commitment and zero-knowledge proof and propose a concept of reliability level to select mobile users. Finally, a general TD algorithm with reliability level is presented to improve the accuracy of the ground truths of sensing tasks. Moreover, theoretical analysis and performance evaluation are conducted, and the evaluation results demonstrate that the PRTD framework outperforms the existing TD frameworks in several evaluation metrics in the synthetic dataset and real-world dataset.

Mobile crowdsensing (MCS) refers to a group of mobile users utilizing their sensing devices to accomplish the same sensing task. However, in vehicular networks, how to evaluate the reliability of sensing vehicles and achieve lightweight privacy preservation are urgent issues. Therefore, this paper proposes a lightweight privacy preservation scheme with efficient reputation management (PPRM) for MCS in vehicular networks. Specifically, we design a lightweight privacy-preserving sensing task matching algorithm which can preserve the location privacy, identity privacy, sensing data privacy, and reputation value privacy while reducing communication and computation overheads of sensing vehicles. In particular, to prevent reputation values from being forged and select reliable sensing vehicles, we present a privacy-preserving reputation value equality verification algorithm to verify reputation values and a privacy-preserving reputation value range proof algorithm to choose sensing vehicles. Afterwards, a three-factor reputation value update algorithm is constructed to efficiently and accurately update the reputation values for sensing vehicles. Simulations are conducted to demonstrate the performance of the PPRM scheme, and the results show that the PPRM scheme significantly outperforms the existing schemes in security and robustness aspects.

Genome-Wide Association Study (GWAS) aims at detecting the association between diseases and Single-Nucleotide Polymorphisms (SNPs) with statistical techniques and has great potential for disease diagnosis. To obtain high-quality results, GWAS requires large-scale genomic data containing individuals’ privacy information. Thus, how to improve the efficiency of GWAS while protecting the privacy of genomic data becomes a critical challenge. In this paper, we propose a secure and efficient GWAS scheme. By using secure three-party computation, we present a series of protocols, i.e., Secure Quality Control, Secure Principle Component Analysis, Secure Cochran-Armitage trend test, and Secure Logistic Regression, to cover the most significant procedures of secure GWAS. In these protocols, a new comparison protocol is designed to reduce communication and improve efficiency. Furthermore, by extending the above comparison protocol to be maliciously secure and utilizing other technologies, e.g., consistency check, we extend the whole GWAS scheme to malicious security with rationally additional overhead. Experimental results demonstrate that our protocols achieve about 33% performance improvement than the state-of-art secure GWAS scheme using two-party computation in terms of runtime and communication in the semi-honest setting. The cost of our scheme in the malicious setting is around 1.5X than that in the semi-honest setting.