Frequency estimation mechanisms are widely applied in domains such as machine learning and cloud computing, where it is desirable to provide statistical information. As a fundamental operation in these domains, frequency estimation utilizes personal data which contains sensitive information while it is necessary to protect sensitive information from others. Motivated by this, we preserve user’s privacy with local differential privacy by obfuscating personal data on the user side. In this paper, we propose frequency estimation mechanisms under utility-optimized local differential privacy (ULDP), which allow the data collector to obtain some non-sensitive values to improve data utility while protecting sensitive values from leaking sensitive information. We propose three frequency estimation mechanisms under $(\epsilon ,\delta )$ -ULDP (uRFM-GRR, uRFM-RAPPOR, uRFM-OLH) to preserve user’s sensitive information. Our proposed mechanisms protect sensitive data with the same privacy guarantee and they are suitable for different scenarios. Besides, in theory, we compare the estimation errors of our proposed mechanisms with existing LDP based mechanisms and show that ours are lower than theirs. Finally, we conduct experiments on synthetic and real-world datasets to evaluate the performance of the three mechanisms. The experimental results demonstrate that our proposed mechanisms are better than the existing LDP based solutions over the same privacy level, while uRFM-OLH frequently performs the best.

In the era of big data, advanced data processing devices and smart sensors greatly benefit us in many areas. As for each individual user, data sharing can be an essential part of the process of data collection and transmission. However, the issue of constant attacks on data privacy arouses huge concerns among the public. This work proposes a personalized federated learning method associated with correlated differential privacy for autonomous driving. First, instead of transmitting raw data to the server following collection, a device that employs federated learning can perform calculations to obtain the training model at each node. Second, we specifically perform a correlated classification analysis to encrypt data that share high relevance, which can minimize the system cost. Then, correlated differential privacy is utilized to achieve the preservation of data privacy before sharing. In contrast to the traditional differential privacy, the proposed solution guarantees enhanced privacy to meet the demands of customization. The experimental results show that our scheme is more refined in terms of user heterogeneity and the utility of data than others without violating privacy.

Utility-optimized local differential privacy (ULDP) considers input domain including non-sensitive values which reduces utility loss by leaking some non-sensitive values, without lowering protection to any sensitive one compared with local differential privacy (LDP). The existing ULDP mechanisms are designed under ϵ-ULDP which preserve sensitive values under ϵ-LDP. Nevertheless, it is still challenging to achieve (ϵ, δ)-ULDP. In this paper, we consider mean aggregation in (ϵ, δ)-ULDP, where sensitive values are protected under (ϵ, δ)-LDP. Specifically, we first propose One-Bit perturbation Mechanism (OBM) for unbiased mean estimation under (ϵ, δ)-LDP and then obtain optimal OBM by minimizing its worst-case error. In OBM, each output is a 1-bit value, and it thus is highly communication-efficient. Second, based on OBM, we design an unbiased mean estimation mechanism in (ϵ, δ)-ULDP, Utility-optimized OBM (UOBM), where sensitive values are strictly protected under (ϵ, δ)-LDP while non-sensitive ones could be disclosed to achieve higher utility. Further, we extend UOBM to the personalized scene where each user has specific privacy budget and sensitive range. Additionally, we theoretically and experimentally compare the proposed mechanisms with existing ones. The results show OBM outperforms existing mechanisms in utility, though its output is just a 1-bit value. UOBM can dramatically decrease the estimation error, compared with OBM.

With the development of smart grid, smart meters play an important role during the process of data collection and delivery, where users’ electricity consumption information may be exposed to malicious users. The attacks to these sensitive information may touch off serious privacy concerns and pose a threat to the security of power system. In this paper, we design a novel framework to satisfy both personalized privacy preservation and customized system cost in the smart metering system, which contains a Rechargeable Battery (RB). First, we adopt differential privacy algorithm to enhance privacy and define the privacy preservation level for the privacy measure. Second, we specially design a comprehensive cost computing system, including electricity cost, battery operation cost, and the privacy budget. Then Markov Decision Process (MDP) is utilized in this work to formulate an optimization problem figuring out the trade-off between privacy and cost. Deploying Sarsa algorithm to address the optimization problem, we can finally obtain an optimal energy management system for different customers according to their personalized demands. Experimental evaluation results show that our system is able to balance privacy protection and cost saving with the introduction of RB.

Privacy-preserving string equality test is a fundamental operation of many algorithms, including privacy-preserving authentication in Internet of Things (IoT). Existing secure equality test schemes can theoretically achieve string equality comparison and preserve the private strings. However, they suffer from heavy computation and communication cost, especially while the strings are of hundreds of bits or longer, which is not suitable for IoT applications. In this paper, we propose an approximate Fast privacy-preserving equality Test Protocol (FTP), which can securely complete string equality test and achieve high running efficiency at the cost of little accuracy loss. We strictly analyze the accuracy of our proposed scheme and formally prove its security. Additionally, we leverage extensive simulation experiments to evaluate the running cost, which confirms our high efficiency; for instance, our proposed FTP can securely compare two 256-bit strings within 0.7 seconds on ordinary laptops.

With the widespread adoption of the Internet of things (IoT) and cloud computing, vehicle intelligent interactive systems have gained significant attention. This has led to extensive research on ensuring privacy protection for in-vehicle communication data. The significance of the privacy study rests on the fact that the leakage of navigation and location information can lead to potential threats to vehicle owners or car manufacturers. Existing differential privacy location protection methods are primarily applied in a continuous query environment, where noise overlay has been reported previously to cause reduced query accuracy. To address this problem, this paper proposes a differential privacy-preserving method based on an M-ary tree based on a k-anonymity set constructed from Hilbert curves. This method reduces the data noise of continuous queries while ensuring the strength of privacy protection. Especially when the privacy budget is 0.5, the proposed algorithm reduces the mean error by about 20000 compared to existing algorithms. M-ary averaging tree structure partitions the anonymized data, improves data utility, and rigorously derives a Laplacian private budget allocation strategy with a small error upper bound. Extensive experiments demonstrate that our proposed method has smaller query errors over a wide range and gradually stabilizes compared to Dwork and EHWT-DP methods with higher efficiency and accuracy.