Federated learning is a promising tool in the Internet-of-Things (IoT) domain for training machine model decentralized manner. Specifically, data owners (e.g., IoT device consumers) keep their raw and only share local computation results to train global of owner an service provider). When executing federated task, contribute communication resources. In this situation, have face privacy issues where attackers may infer property or recover based on shared information. Considering these...

Federated learning (FL) through its novel applications and services has enhanced presence as a promising tool in the Internet of Things (IoT) domain. Specifically, multiaccess edge computing setup with host IoT devices, FL is most suitable since it leverages distributed client data to train high-performance deep (DL) models while keeping private. However, underlying neural networks (DNNs) are huge, preventing direct deployment onto resource-constrained memory-limited devices. Besides,...

The transportation network company (TNC) services efficiently pair the passengers with vehicles/drivers through mobile applications, such as Uber, Lyft, Didi, etc. TNC definitely facilitate traveling of passengers, while it is equally important to effectively and intelligently schedule routes cruising vehicles improve drivers' revenues. From side, most critical question address how reduce time, efficiency/earnings by using their own provide services. In this paper, we propose a deep...

Deep learning holds a great promise of revolutionizing healthcare and medicine. Unfortunately, various inference attack models demonstrated that deep puts sensitive patient information at risk. The high capacity neural networks is the main reason behind privacy loss. In particular, in training data can be unintentionally memorized by network. Adversarial parties extract given ability to access or query this paper, we propose novel privacy-preserving mechanism for networks. Our approach adds...

Machine learning is increasingly becoming a powerful tool to make decisions in wide variety of applications, such as medical diagnosis and autonomous driving. Privacy concerns related the training data unfair behaviors some with regard certain attributes (e.g., sex, race) are more critical. Thus, constructing fair machine model while simultaneously providing privacy protection becomes challenging problem. In this paper, we focus on design classification fairness differential guarantees by...

Indoor location-based services (LBS) are widely used in large-scale indoor buildings, such as high-rise hospitals and multi-story shopping malls. At the same time, location privacy protection three-dimensional (3D) space has recently attracted considerable attention. Currently, most existing schemes focus on two-dimensional (2D) fail to prevent inference attacks when user's data include height dimension, i.e., 3D geolocation. Enlightened by concept of differential privacy, this paper we...

Objective To use the United States National Health and Nutrition Examination Study (NHANES) to develop validate a risk-prediction nomogram for cognitive impairment in people aged over 60 years. Methods A total of 2,802 participants (aged ≥ years) from NHANES were analyzed. The least absolute shrinkage selection operator (LASSO) regression model multivariable logistic analysis used variable development. ROC-AUC, calibration curve, decision curve (DCA) evaluate nomogram’s performance. Results...

The rapid development of indoor location-based services (LBS) has raised concerns about location privacy protection in the 3-dimensional (3D) space. existing 2-dimensional (2D) mechanisms (LPPMs) cannot effectively resist attacks 3D environments. Furthermore, users may have various sensitive attributes at different locations and times. In this paper, we first formally study relationship between two complementary notions geo-indistinguishability distortion (i.e., expected inference error)...

To embrace the era of big data, there has been growing interest in designing distributed machine learning to exploit collective computing power local nodes. Alternating Direction Method Multipliers (ADMM) is one most popular methods. This method applies iterative computations over datasets at each agent and computation results exchange between neighbors. During this process, data privacy leakage arises when performing sensitive data. Although many differentially private ADMM algorithms have...

Deep learning has attracted broad interest in healthcare and medical communities. However, there been little research into the privacy issues created by deep networks trained for applications. Recently developed inference attack algorithms indicate that images text records can be reconstructed malicious parties have ability to query networks. This gives rise concern electronic health containing sensitive patient information are vulnerable these attacks. paper aims attract from researchers...

Collaborative learning has received huge interests due to its capability of exploiting the collective computing power wireless edge devices. However, during process, model updates using local private samples and large-scale parameter exchanges among agents impose severe privacy concerns communication bottleneck. In this paper, address these problems, we propose two differentially (DP) efficient algorithms, called Q-DPSGD-1 Q-DPSGD-2. Q-DPSGD-1, each agent first performs by a DP gradient...

The advance in mobile communications has escalated the use of transportation network company (TNC) services by residents. TNC efficiently pair passengers with vehicles/drivers through applications such as Uber, Lyft, Didi, etc. definitely facilitate traveling passengers, while it is equally important to effectively and intelligently schedule routes cruising vehicles improve drivers' revenues. From side, most critical question address how reduce time, efficiency/earnings using their own...

Machine learning has a vast outreach in principal applications and uses large amount of data to train the models, prompting viable easy use Learning as Service (MLaaS). This flexible paradigm however, could have immense privacy implications since training often contains sensitive features, adversarial access such models pose security risk. In attacks model inversion attack on system used for face recognition, an adversary output (target label) reconstruct input (image target individual from...

With the wide range application of machine learning in healthcare for helping humans drive crucial decisions, data privacy becomes an inevitable concern due to utilization sensitive such as patients records and registers a company. Thus, constructing preserving model while still maintaining high accuracy challenging problem. In this article, we propose two differentially private algorithms, i.e., Output Perturbation with aGM (OPERA) Gradient (GRPUA) empirical risk minimization, useful method...

As a promising method of central model training on decentralized device data while securing user privacy, Federated Learning (FL) is becoming popular in the Internet Things (IoT) design. However, when collected by IoT devices are highly skewed non-independent and identically distributed (non-IID) manner, accuracy vanilla FL cannot be guaranteed. Although there exist various solutions that try to address bottleneck with non-IID data, most them suffer from extra intolerable communication...

In recent years, generative adversarial network (GAN) has attracted great attention due to its impressive performance and potential numerous applications, such as data augmentation, real-like image synthesis, compression improvement, etc. The generator in GAN learns the density of distribution from real order generate high fidelity fake samples latent space deceive discriminator. Despite advantages, can easily memorize training because model complexity deep neural networks. Thus, a with...