Differential Privacy Preserving Research Articles

Child Health Dataset Publishing and Mining Based on Differential Privacy Preservation

With the emergence and development of application requirements such as data analysis and publishing, it is particularly important to use differential privacy protection technology to provide more reliable, secure, and compliant datasets for research in the field of children’s health. This paper focuses on the differential privacy protection of the ultrasound examination health dataset of adolescents in southern Texas from three aspects: differential privacy protection with output perturbation on basic statistics, publication of differential privacy marginal histogram and synthesized data, and a machine learning differential privacy learning algorithm. Firstly, differential privacy protection results with output perturbation show that Laplace and Gaussian mechanisms for numerical data, as well as the exponential mechanism for non-numerical data, can achieve the goal of protecting privacy. The exponential mechanism provides higher privacy protection. Secondly, a differential privacy marginal histogram with four attributes can be obtained with an appropriate privacy budget that approximates the marginal histogram of the original data. In order to publish synthetic data, we construct a synthetic query to obtain the corresponding differential privacy histogram for two attributes. Further, a synthetic dataset can be constructed by following the data distribution of the original dataset and the quality of the synthetic data publication can also be evaluated by the mean square error and error rate. Finally, consider a differential privacy logistic regression model under machine learning to predict whether children have fatty liver in binary classification tasks. The experimental results show that the model combined with quadratic perturbation has better accuracy and privacy protection. This paper can provide differential privacy protection models under different demands, which provides important data release and analysis options for data managers and research organizations, in addition to enriching the research on child health data releasing and mining.

Full Differential Privacy Preserving for Switched LPV Systems

Full Differential Privacy Preserving for Switched LPV Systems

Energy Efficient and Differentially Private Federated Learning via a Piggyback Approach

This paper aims to develop a differential private federated learning (FL) scheme with the least artificial noises added while minimizing the energy consumption of participating mobile devices. By observing that some communication efficient FL approaches and even the nature of wireless communications contribute to the differential privacy (DP) preservation of training data on mobile devices, in this paper, we propose to jointly leverage gradient compression techniques (i.e., gradient quantization and sparsification) and additive white Gaussian noises (AWGN) in wireless channels to develop a piggyback DP approach for FL over mobile devices. Even with the piggyback DP approach, information distortion caused by gradient compression and noise perturbation may slow down FL convergence, which in turn consumes more energy of mobile devices for local computing and model update communications. Thus, we theoretically analyze FL convergence and formulate an energy efficient FL optimization under piggyback DP, transmission power, and FL convergence constraints. Furthermore, we propose an efficient iterative algorithm where closed-form solutions for artificial DP noise and power control are derived. Extensive simulation and experimental results demonstrate the effectiveness of the proposed scheme in terms of energy efficiency and privacy preservation.

Differential Privacy Preservation for Continuous Release of Real-Time Location Data.

Continuous real-time location data is very important in the big data era, but the privacy issues involved is also a considerable topic. It is not only necessary to protect the location privacy at each release moment, but also have to consider the impact of data correlation. Correlated Laplace Mechanism (CLM) is a sophisticated method to implement differential privacy on correlated time series. This paper aims to solve the key problems of applying CLM in continuous location release. Based on the finding that the location increment is approximately stationary in many scenarios, a location correlation estimation method based on the location increment is proposed to solve the problem of nonstationary location data correlation estimation; an adaptive adjustment model for the CLM filter based on parameter quantization idea (QCLM) as well as its effective implementation named QCLM-Lowpass utilizing the lowpass spectral characteristics of location data series is proposed to solve the problem of output deviations due to the undesired transient response of the CLM filter in time-varying environments. Extensive simulations and real data experiments validate the effectiveness of the proposed approach and show that the privacy scheme based on QCLM-Lowpass can offer a better balance between the ability to resist correlation-based attacks and data availability.

Preserving Differential Privacy in Deep Learning Based on Feature Relevance Region Segmentation

In the era of Big Data, deep learning techniques provide intelligent solutions for various problems in real-life scenarios. However, deep neural networks depend on large-scale datasets including sensitive data, which causes the potential risk of privacy leakage. In addition, various constantly evolving attack methods are also threatening the data security in deep learning models. Protecting data privacy effectively at a lower cost has become an urgent challenge. This paper proposes an Adaptive Feature Relevance Region Segmentation (AFRRS) mechanism to provide differential privacy preservation. The core idea is to divide the input features into different regions with different relevance according to the relevance between input features and the model output. Less noise is intentionally injected into the region with stronger relevance, and more noise is injected into the regions with weaker relevance. Furthermore, we perturb loss functions by injecting noise into the polynomial coefficients of the expansion of the objective function to protect the privacy of data labels. Theoretical analysis and experiments have shown that the proposed AFRRS mechanism can not only provide strong privacy preservation for the deep learning model, but also maintain the good utility of the model under a given moderate privacy budget compared with existing methods.

Differentially Private Distributed Optimization With an Event-Triggered Mechanism

This study concentrates on the differential private distributed optimization problem with an event-triggered mechanism, whose goals include preserving the privacy of agents’ initial states and local cost functions and improving communication efficiency. A distributed event-triggered mechanism is integrated into the differentially private subgradient-push distributed optimization algorithm and then a new algorithm named as DP-ETSP is designed, where the real-time information propagation among agents is avoided. Additionally, under the proposed event-triggered mechanism, an analysis of mean-square consensus and optimality over time-varying directed networks is made when the added Laplace noises meet some specific decaying conditions. Convergence rate results are further established under a specific stepsize, which are equal to the rate of stochastic gradient-push algorithm without event-triggered communication. Moreover, the differential privacy preservation performance is analyzed and the rule for selecting privacy level is discussed. Finally, the feasibility and effectiveness of DP-ETSP are verified in two simulation cases.

SeRaNDiP

Personal data collected from today's wearable sensors contain a rich amount of information that can reveal a user's identity. Differential privacy (DP) is a well-known technique for protecting the privacy of the sensor data being sent to community sensing applications while preserving its statistical properties. However, differential privacy algorithms are computationally expensive, requiring user-level random noise generation which incurs high overheads on wearables with constrained hardware resources. In this paper, we propose SeRaNDiP -- which utilizes the inherent random noise existing in wearable sensors for distributed differential privacy. We show how various hardware configuration parameters available in wearable sensors can enable different amounts of inherent sensor noise and ensure distributed differential privacy guarantee for various community sensing applications with varying sizes of populations. Our evaluations of SeRaNDiP on five wearable sensors that are widely used in today's commercial wearables -- MPU-9250 accelerometer, ADXL345 accelerometer, BMP 388 barometer, MLP 3115A2 barometer, and MLX90632 body temperature sensor show a 1.4X-1.8X computation/communication speedup and 1.2X-1.5X energy savings against state-of-the-art DP implementation. To the best of our knowledge, SeRaNDiP is the first framework to leverage the inherent random sensor noise for differential privacy preservation in community sensing without any hardware modification.

DPPT : A differential privacy preservation technique for cyber–physical system

Nowadays Cyber–Physical System (CPS) is implemented in several areas, such as health care, agriculture, transportation, etc. The main aim of CPS is for automation or minimization of human interaction. The conventional Machine Learning (ML) based approach involves sensitive issues such as information leakage. That issue is resolved using the Federated Learning (FL) based approach partially. The FL uses a decentralized environment for training the models, where models are trained locally, and only the updated part is sent to the data center or cloud server. It improves data privacy as the clients send only the local updates (model gradients). But also there is a high risk of data leakage while publishing the data in FL based System. A Differential Privacy Preservation Technique (DPPT) is an efficient technique to preserve information leakage. In this paper, a differential privacy preservation technique has been proposed by adding random noise to a data sample in order to generate anonymity. The performance of the model, and data quality are evaluated as privacy preservation techniques often degrade data quality.

Differential Privacy Preserving Using TensorFlow DP-SGD and 2D-CNN for Large-Scale Image Data

Abstract: Although smart wearables have many potential advantages, their widespread and ongoing use raises a number of privacy issues and difficult information security challenges. This article, present a thorough analysis of current wearable sensorbased big data analytics applications that protect user privacy. We draw attention to the fundamental aspects of privacy and security for applications on wearable technology. Then, we look at how deep learning techniques like 2D CNN are used for better evaluation and privacy preservation as well as for differential privacy of Tensor flow. DP-SGD (Differentially private stochastic gradient descent)The metrics are epsilon as well as accuracy, with 0.56 epsilon and 85.17% accuracy for three epochs and 100.09 epsilon and 95.28 accuracies for twenty epochs, respectively. Model training Accuracy is 94.71%. Also, present a case study on privacy-preserving machine learning techniques. Herein, we theoretically and empirically evaluate the privacypreserving deep learning framework's performance. We explain the implementation details of a case study of a secure prediction service using the 2D convolutional neural network (2D CNN) model

Differential Privacy Preservation for Location Semantics.

With the rapid development of intelligent mobile terminals and communication technologies, location-based services (LBSs) have become an essential part of users' lives. LBS providers upload and share the collected users' location data. The more commonly used methods for location privacy protection are differential privacy and its extensions. However, the semantic information about location, which is an integral part of the location data, often contains sensitive user information. Most existing research methods have failed to pay enough attention to protecting the semantic information in the location data. To remedy this problem, two different scenarios for location semantic privacy protection methods are proposed in this paper to address single-point and continuous location queries. Simulation experiments on real social location check-in datasets, and comparison of three different privacy protection mechanisms, show that our solution demonstrates good service quality and privacy protection considering location semantics.

DP-AGM: A Differential Privacy Preserving Method for Binary Relationship in Mobile Networks

DP-AGM: A Differential Privacy Preserving Method for Binary Relationship in Mobile Networks

Data-Driven Transportation Network Company Vehicle Scheduling With Users’ Location Differential Privacy Preservation

With the popularity of mobile devices with global positioning system (GPS), transportation network company (TNC) service has become an indispensable option of people's daily commute. However, it also provides opportunities for malicious parties to compromise TNC users’ location privacy. There are great challenges to preserve TNC users’ location privacy while improving the revenue of TNC and its quality of service (QoS). To address this issue, we propose a novel scheme to schedule the TNC vehicles while preserving the TNC users’ location differential privacy. Briefly, we add high dimensional Laplace noises to guarantee the TNC users’ geo-indistinguishability. Due to the differential private obfuscation, the demand for TNC vehicles in an area becomes uncertain. Thus, we employ the data-driven approach to characterize users’ demand uncertainty, formulate the TNC's revenue maximization problem into risk-averse stochastic programming, and provide corresponding feasible solutions. Using the released public data of Didi Chuxing, we conduct extensive simulations to evaluate the performance of the proposed scheduling scheme and compare the results under different <inline-formula><tex-math notation="LaTeX">$\zeta$</tex-math></inline-formula> -structure metrics. The results show that the proposed scheme can efficiently schedule the TNC vehicles, maximize the TNC's revenue and provide a better service for TNC users while protecting the TNC users’ location privacy.

A Density Peaking Clustering Algorithm for Differential Privacy Preservation

A Density Peaking Clustering Algorithm for Differential Privacy Preservation

Differential privacy preservation for graph auto-encoders: A novel anonymous graph publishing model

In recent years, the number of users in social networks has grown substantially, and more data-intensive applications have been developed. This creates a demand for the ability to mine large-scale graph data more efficiently, so that the information mined can be maximized (e.g., mining social relationships between people). However, the direct publication of the original graphs leads to potential leakage of users’ privacy. Therefore, graph anonymization techniques are often utilized to process the original graphs. A key challenge of it lies in the balance between anonymity and usability. In this paper, we introduced the idea of graph auto-encoder, a fundamental element in graph neural networks, and proposed the Differential Privacy Deep Graph Auto-Encoder (DP-DGAE). Our main idea is to convert the anonymous graph publishing problem into a privacy-preserving problem for generative models, and optimize the models in terms of both privacy and usability using a multi-task learning approach. Theoretical analysis and experimental evaluations show that the DP-DGAE achieves anonymity while ensuring usability.

Augmentation-Based Edge Differentially Private Path Publishing in Networks

Paths in a given network represent the occurrence sequences of nodes in many real world applications, such as disease transmission chains, object trajectories and data access sequences. In this paper, we address the problem of publishing edge-privacy preserved path information for a single path such that legitimate users with the full knowledge of the network can reconstruct the path with the published information, but not adversaries, even if they have the maximum background knowledge of all the vertices and all edges but one (on the path) of the network. Existing studies on edge privacy against inference attacks focus on publishing either differential privacy (DP) noise injected graph statistics or DP edge perturbed graph topology to achieve edge differential privacy preservation. However, none of them provides an assurance on both edge privacy and data utility. To effectively protect edge privacy and maintain data utility, we propose a novel scheme of DP augmentation instead of DP perturbation as did in existing work, that publishes a simple-topology graph containing an augmented path with fake edges and vertices applying differential privacy to protect the actual path, such that only the legitimate users are able to reconstruct the actual path with high probability. We theoretically analyse the performance of our algorithm in terms of output quality on differential privacy and utility, and execution efficiency. We also conduct extensive experimental evaluations on a high-performance cluster system to validate our analytical results.

Enhanced Privacy Preserving for Social Networks Relational Data Based on Personalized Differential Privacy

With the popularization and development of social software, more and more people join the social network, which produces a lot of valuable information, but also contains plenty of sensitive privacy information. To achieve the personalized privacy protection of massive social network relational data, a privacy enhancement method for social networks relational data based on personalized differential privacy is proposed. And a dimensionality reduction segmentation sampling (DRS-S) algorithm is proposed to implement this method. First, in order to solve the problem of inefficiency caused by the excessive amount of data in social networks, dimension reduction and segmentation are carried out to divide the data into groups. According to the privacy protection requirements of different users, we adopt sampling method to protect users with different privacy requirements at different levels, so as to realize personalized different privacy. After that, the noise is added to the protected data to satisfy the privacy budget. Then publish the social network data. Finally, the proposed algorithm is compared with the traditional personalized differential privacy (PDP) algorithm and privacy preserving approach based on clustering and noise (PBCN) in real data set, the experimental results demonstrate that the quality of privacy protection and data availability of DRS-S are better than that of PDP algorithm and PBCN algorithm.

A Novel Distributed Differential Privacy Preserving Based on Random Forest in Data Centers

Traditional privacy-preserving technologies have been unable to provide adequate protection and are vulnerable to background knowledge. MapReduce is parallel distributed computing model which has the advantages of good scalability and high fault tolerance. In this paper, we design a differential privacy protection based on MapReduce for the security problems faced in the distributed environment. Firstly, we propose a random forest algorithm DPMRRF that satisfies differential privacy under the MapReduce framework. Secondly, we use the MapReduce framework in the Hadoop platform to make the classification results satisfy differential privacy by adding random noise values to the leaf nodes. Finally, through the design of the feature selection scheme combined with the index mechanism, three different privacy budget allocation methods are adopted, namely, equal distribution, equal distribution and equal distribution. Experiments show that the algorithm has better classification accuracy while ensuring data availability. The proposed algorithm effectively improves the classification accuracy and operating efficiency, and greatly reduces the amount of computation.

Differential Privacy Preserving in Big data Analytics for Body Area Networks

Abstract: Body Area Networks (BANs), collects enormous data by wearable sensors which contain sensitive information such as physical condition, location information, and so on, which needs protection. Preservation of privacy in big data has emerged as an absolute prerequisite for exchanging private data in terms of data analysis, validation, and publishing. Previous methods and traditional methods like k-anonymity and other anonymization techniques have overlooked privacy protection issues resulting to privacy infringement. In this work, a differential privacy protection scheme for ‘big data in body area network’ is developed. Compared with previous methods, the proposed privacy protection scheme is best in terms of availability and reliability. Exploratory results demonstrate that, even when the attacker has full background knowledge, the proposed scheme can still provide enough interference to big sensitive data so as to preserve the privacy. Keywords: BAN’s, Privacy, Differential Privacy, Noisy response

An adaptive federated learning scheme with differential privacy preserving

Driven by the upcoming development of the sixth-generation communication system (6G), the distributed machine learning schemes represented by federated learning has shown advantages in data utilization and multi-party cooperative model training. The total communication costs of federated learning is related to the number of communication rounds, the communication consumption of each participants, the setting of reasonable learning rate and the guarantee of calculation fairness. In addition, the isolating data strategy in the federated learning framework cannot completely guarantee the privacy security of users. Motivated by the above problems, this paper proposes a federated learning scheme combined with the adaptive gradient descent strategy and differential privacy mechanism, which is suitable for multi-party collaborative modeling scenarios. To ensure that federated learning scheme can train efficiently with limited communications costs, the adaptive learning rate algorithm is innovatively used to adjust the gradient descent process and avoid the model overfitting and fluctuation phenomena, so as to improve the modeling efficiency and performance in multi-party calculation scenarios. Furthermore, in order to adapt to the ultra-large-scale distributed secure computing scenario, this research introduces differential privacy mechanism to resist various background knowledge attacks. Experimental results demonstrate that the proposed adaptive federated learning model performs better than the traditional models under fixed communication costs. This novel modeling scheme also has strong robustness to different super-parameter settings and provides stronger quantifiable privacy preserving for federated learning process.

Adaptive differential privacy preserving based on multi‐objective optimization in deep neural networks

SummaryPrivacy data security has become an important bottleneck for the overall development of artificial intelligence and a key challenge that needs to be broken in the Internet era. The current research mainly considers differential privacy to effectively protect the private information in the data. However, as the noise increases, the precision of the training model will decrease. In order to solve above problem, an adaptive differential privacy (ADP) method is constructed and applied to deep neural networks. ADP adds noise adaptively in the training process according to the importance of features. We also build the differential privacy multi‐objective optimization model (DPMOM). DPMOM adopts multi‐objective optimization characteristics, takes accuracy and privacy protection as the optimization objectives. It optimizes the super parameters of deep neural networks and the noise of differential privacy. In addition, to better solve the ADP model, with the NSGA‐II algorithm as the basic framework, a multi‐objective optimization algorithm based on differential privacy protection (DPPMOA) is designed. Simulation experiments show that compared with other machine learning methods and differentially private stochastic gradient descent, the accuracy of ADP is higher under the same amount of noise. Through comparison with NSGA‐II, IBEA, PESA‐II, and AGE‐II, DPPMOA is proved that the solution set of this algorithm is better.