Since traditional federated learning algorithms cannot provide sufficient privacy guarantees, an increasing number of approaches apply local differential privacy (LDP) techniques to federated learning to provide strict privacy guarantees. However, the privacy budget heavily increases proportionally with the dimension of the parameters, and the large variance generated by the perturbation mechanisms leads to poor performance of the final model. In this paper, we propose a novel privacy-preserving edge federated learning framework based on LDP (PPeFL). Specifically, we present three LDP mechanisms to address the privacy problems in the federated learning process. The proposed filtering and screening with exponential mechanism (FS-EM) filters out the better parameters for global aggregation based on the contribution of weight parameters to the neural network. Thus, we can not only solve the problem of fast growth of privacy budget when applying perturbation mechanism locally, but also greatly reduce the communication costs. In addition, the proposed data perturbation mechanism with stronger privacy (DPM-SP) allows a secondary scrambling of the original data of participants and can provide strong security. Further, a data perturbation mechanism with enhanced-utility (DPM-EU) is proposed in order to reduce the variance introduced by the perturbation. Finally, extensive experiments are performed to illustrate that the PPeFL scheme is practical and efficient, providing stronger privacy protection while ensuring utility.

With the rapid development of machine learning in the medical cloud system, cloud-assisted medical computing provides a concrete platform for remote rapid medical diagnosis services. Support vector machine (SVM), as one of the important algorithms of machine learning, has been widely used in the field of medical diagnosis for its high classification accuracy and efficiency. In some existing schemes, healthcare providers train diagnostic models with SVM algorithms and provide online diagnostic services to doctors. Doctors send the patient’s case report to the diagnostic models to obtain the results and assist in clinical diagnosis. However, case report involves patients’ privacy, and patients do not want their sensitive information to be leaked. Therefore, the protection of patient’s privacy has become an important research direction in the field of online medical diagnosis. In this paper, we propose a privacy-preserving medical diagnosis scheme based on multi-class SVMs. The scheme is based on the distributed two trapdoors public key cryptosystem (DT-PKC) and Boneh-Goh-Nissim (BGN) cryptosystem. We design a secure computing protocol to compute the core process of the SVM classification algorithm. Our scheme can deal with both linearly separable data and nonlinear data while protecting the privacy of user data and support vectors. The results show that our scheme is secure, reliable, scalable with high accuracy.

As a popular machine learning method, federated learning (FL) can effectively solve the issues of data silos and data privacy. However, traditional federated learning schemes cannot provide sufficient privacy protection. Furthermore, most secure federated learning schemes based on local differential privacy (LDP) ignore an important issue: they do not consider each client’s differentiated privacy requirements. This paper introduces a perturbation algorithm (PDPM) that satisfies personalized local differential privacy (PLDP), resolving the issue of inadequate or excessive privacy protection for some participants due to the same privacy budget set for all clients. The algorithm enables clients to adjust the privacy parameters according to the sensitivity of their data, thus allowing the scheme to provide personalized privacy protection. To ensure the privacy of the scheme, we have conducted a strict privacy proof and simulated the scheme on both synthetic and real data sets. Experiments have demonstrated that our scheme is successful in producing high-quality models and fulfilling the demands of personalized privacy protection.

Internet of Things has remarkable effects in human’s daily life. It is important for users and sensors to securely access data collected by low-cost sensors via Internet in real-time IoT applications. There exist many authentication protocols for guaranteeing secure communication between users and sensors. However, in some protocols, the privacy of unattended sensors subjected to capture node attacks cannot be guaranteed. Moreover, the sensors subjected to physical tampering attacks can still execute normally the authentication process. Besides, an authentication protocol should be lightweight due to the restricted computing power and storage of the sensors. The idea of designing a more secure and lightweight authentication protocol engender this article. The proposed protocol can provide the physical security through physically unclonable function (PUF), require no additional phase to update challenge-response pairs (CRPs), and store a single CRP for each sensor. At the same time, the proposed protocol utilizes three factors, such as personal biometrics, smartcard and password, to strengthen the security contrasting with two factors, and manipulates some basic cryptographic operations, including bitwise-exclusive-OR (XOR) and hash function, to achieve the lightweight performance. Moreover, both formal security analysis based on Real-Or-Random (ROR) and informal security analysis demonstrate the security of the proposed protocol. Compared with the existing related protocols, the proposed protocol has the advantage in terms of security, functionality and computation costs. Finally, a NS3 simulation on measuring various network performance parameters indicates that the proposed protocol is practical in IoT environment.

Fully homomorphic encryption supports meaningful computations on encrypted data, and hence, is widely used in cloud computing and big data environments. Recently, Li et al. constructed an efficient symmetric fully homomorphic encryption scheme and utilized it to design a privacy-preserving-outsourced association rule mining scheme. Their proposal allows multiple data owners to jointly mine some association rules without sacrificing the data privacy. The security of the homomorphic encryption scheme against the known-plaintext attacks was established by examining the hardness of solving nonlinear systems. However, in this paper, we illustrate that the security of Li et al.'s homomorphic encryption is overvalued. First, we show that we can recover the first part of the secret key from several known plaintext/ciphertext pairs with the continued fraction algorithm. Second, we find that we can retrieve the second part of the secret key through the Euclidean algorithm for the greatest common divisor problem. Experiments on the suggested parameters demonstrate that in case of more than two homomorphic multiplications, all the secret keys of the randomly instantiated Li et al.'s encryption schemes can be very efficiently recovered, and the success probability is at least 98% for one homomorphic multiplication.

Artificial intelligence (AI) has demonstrated huge potential in a variety of real-world applications. However, some significant considerations like fairness, transparency and trustworthiness are still challenging when applying AI to trust-oriented applications such as E-voting. E-voting plays a significant role in democratic societies, which requires voters and initiators have strong mutual trust. In this paper, we aim to facilitate the consolidation of AI ecosystems by developing a blockchain-based traceable self-tallying e-voting system. We take advantage of an event-oriented linkable group signature and a homomorphic time-lock puzzle to balance the anonymity and accountability, and the voting scale and efficiency of an e-voting system. The proposed e-voting protocol supports additional functions like multi-choice and self-tallying. We prove that the proposed protocol satisfies anonymity, time-bounded privacy, linkability and full-traceability. We also evaluate the time cost of off-chain operations and the gas cost of on-chain operations, which show the proposed e-voting protocol is practical and can be adopted in real-world applications.