Nonconvex quadratically constrained programs (QCPs) are generally NP-hard and challenging problems. In this paper, we propose two novel mixed-integer linear programming (MILP) approximations for a nonconvex QCP. Our method begins by utilizing an eigenvalue-based decomposition to express the nonconvex quadratic function as the difference of two convex functions. We then introduce an additional variable to partition each nonconvex constraint into a second-order cone (SOC) constraint and the complement of an SOC constraint. We employ two polyhedral approximation approaches to approximate the SOC constraint. The complement of an SOC constraint is approximated using a combination of linear and complementarity constraints. As a result, we approximate the nonconvex QCP with two linear programs with complementarity constraints (LPCCs). More importantly, we prove that the optimal values of the LPCCs asymptotically converge to that of the original nonconvex QCP. By proving the boundedness of the LPCCs, we further reformulate the LPCCs as MILPs. We demonstrate the effectiveness of our approaches via numerical experiments by applying our proposed approximations to randomly generated instances and two application problems: the joint decision and estimation problem and the two-trust-region subproblem. The numerical results show significant advantages of our approaches in terms of solution quality and computational time compared with existing benchmark approaches. History: Accepted by Pascal Van Hentenryck, Area Editor for Computational Modeling: Methods & Analysis. Funding: K. Pan was supported in part by the Research Grants Council of Hong Kong [Grant 15503723]. J. Cheng and B. Yang were supported in part by the Office of Naval Research [Grant N00014-20-1-2154]. J. Cheng was supported in part by the National Science Foundation [Grant ECCS-2404412]. B. Yang was supported in part by the Air Force Office of Scientific Research [Grant FA9550-23-1-0508]. Supplemental Material: The software that supports the findings of this study is available within the paper and its Supplemental Information ( https://pubsonline.informs.org/doi/suppl/10.1287/ijoc.2024.0719 ) as well as from the IJOC GitHub software repository ( https://github.com/INFORMSJoC/2024.0719 ). The complete IJOC Software and Data Repository is available at https://informsjoc.github.io/ .

The increasing need for sharing healthcare data and collaborating on clinical research has raised privacy concerns. Health information leakage due to malicious attacks can lead to serious problems such as misdiagnoses and patient identification issues. Privacy-preserving machine learning (PPML) and privacy-enhancing technologies, particularly federated learning (FL), have emerged in recent years as innovative solutions to balance privacy protection with data utility; however, they also suffer from inherent privacy vulnerabilities. Model inversion attacks constitute major threats to data sharing in federated learning. Researchers have proposed many defenses against model inversion attacks. However, current defense methods for healthcare data lack generalizability, i.e., existing solutions may not be applicable to data from a broader range of populations. In addition, most existing defense methods are tested using non-healthcare data, which raises concerns about their applicability to real-world healthcare systems. In this study, we present a defense against model inversion attacks in federated learning. We achieve this using latent data perturbation and minimax optimization, utilizing both general and medical image datasets. We compare our method against two baselines and observe a reduction of at least 4% in the attacker’s accuracy when classifying reconstructed images, while maintaining model utility within a 1.5% drop of client classification accuracy of the undefended model. These results demonstrate improved privacy protection with minimal utility loss and suggest the potential for a generalizable defense in healthcare settings.

Conventional stress monitoring relies on episodic, symptom-focused interventions, missing the need for continuous, accessible, and cost-efficient solutions. State-of-the-art approaches use rigid, silicon-based wearables, which, though capable of multitasking, are not optimized for lightweight, flexible wear, limiting their practicality for continuous monitoring. In contrast, flexible electronics (FE) offer flexibility and low manufacturing costs, enabling real-time stress monitoring circuits. However, implementing complex circuits like machine learning (ML) classifiers in FE is challenging due to integration and power constraints. Previous research has explored flexible biosensors and ADCs, but classifier design for stress detection remains underexplored. This work presents the first comprehensive design space exploration of low-power, flexible stress classifiers. We cover various ML classifiers, feature selection, and neural simplification algorithms, with over 1200 flexible classifiers. To optimize hardware efficiency, fully customized circuits with low-precision arithmetic are designed in each case. Our exploration provides insights into designing real-time stress classifiers that offer higher accuracy than current methods, while being low-cost, conformable, and ensuring low power and compact size.

Stress, recognized widely as a substantial health concern, adversely affects individuals by undermining both their physical and mental well being. Prior studies on stress monitoring and management utilize a centralized cloud-based approach that combines data from each client for modeling. However, such a centralized approach raises data privacy concerns. To preserve privacy, decentralized federated learning (FL) has been proposed as a potential alternative framework. Nevertheless, existing FL algorithms have to deal with data heterogeneity; data skewness in each participant can significantly degrade the overall model performance. To tackle this challenge, we present a personalized, low-overhead clustered FL algorithm for stress-level recognition. The proposed algorithm outperforms two state-of-the-art baseline algorithms by providing over 7% and 12% increase in accuracy, respectively. The proposed algorithm also obtains a reduction of 37.5% and 9.6% in the training runtime compared to the two baseline algorithms. We also present a novel cold-start algorithm for new clients who join the trained system. Our results suggest that this cold-start algorithm is robust in terms of individual classification accuracy and total training time.

OBJECTIVE: Electronic health records (EHRs) are rich sources of patient-level data, offering valuable resources for medical data analysis. However, privacy concerns often restrict access to EHRs, hindering downstream analysis. Current EHR deidentification methods are flawed and can lead to potential privacy leakage. Additionally, existing publicly available EHR databases are limited, preventing the advancement of medical research using EHR. This study aims to overcome these challenges by generating realistic and privacy-preserving synthetic EHRs time series efficiently. MATERIALS AND METHODS: We introduce a new method for generating diverse and realistic synthetic EHR time series data using denoizing diffusion probabilistic models. We conducted experiments on 6 databases: Medical Information Mart for Intensive Care III and IV, the eICU Collaborative Research Database (eICU), and non-EHR datasets on Stocks and Energy. We compared our proposed method with 8 existing methods. RESULTS: Our results demonstrate that our approach significantly outperforms all existing methods in terms of data fidelity while requiring less training effort. Additionally, data generated by our method yield a lower discriminative accuracy compared to other baseline methods, indicating the proposed method can generate data with less privacy risk. DISCUSSION: The proposed model utilizes a mixed diffusion process to generate realistic synthetic EHR samples that protect patient privacy. This method could be useful in tackling data availability issues in the field of healthcare by reducing barrier to EHR access and supporting research in machine learning for health. CONCLUSION: The proposed diffusion model-based method can reliably and efficiently generate synthetic EHR time series, which facilitates the downstream medical data analysis. Our numerical results show the superiority of the proposed method over all other existing methods.

The digital transformation is characterized by the convergence of technologies—from the Internet of Things (IoT) to edge–fog–cloud computing, artificial intelligence (AI), and Blockchain—in multiple dimensions, blurring the lines between the physical and digital worlds. Although these innovations have evolved independently over time, they are increasingly becoming more intertwined, driving the development of new business models. With more adaptation, embracement, and development, we are witnessing a steady convergence and fusion of these technologies resulting in an unprecedented paradigm shift that is expected to disrupt and reshape the next-generation systems in vertical domains in a way that the capabilities of the technologies are aligned in the best possible way to complement each other. Despite the fact that the convergence of the four technologies can potentially tackle the main shortcomings of the existing systems, its adoption is still in its infancy phase, suffering from several issues, such as the absence of consensus toward any reference models or best practices. This article provides a comprehensive insight into the fusions of these paradigms by discussing a blend of topics addressing all the importation aspects from design to deployment. We will begin this article by providing an in-depth discussion on the main requirements, state-of-the-art reference architectures, applications, and challenges. Following this, we will present a reference architecture and a case study on privacy-preserving stress monitoring and management to better elaborate on the corresponding details and considerations.

The conventional mental healthcare regime often follows a symptom-focused and episodic approach in a noncontinuous manner, wherein the individual discretely records their biomarker levels or vital signs for a short period prior to a subsequent doctor’s visit. Recognizing that each individual is unique and requires continuous stress monitoring and personally tailored treatment, we propose a holistic hybrid edge–cloud Wearable Internet of Things (WIoT)-based online stress monitoring solution to address the above needs. To eliminate the latency associated with cloud access, appropriate edge models—spiking neural network (SNN), Conditionally Parameterized Convolutions (CondConv), and support vector machine (SVM)—are trained, enabling low-energy real-time stress assessment near the subjects on the spot. This work leverages design-space exploration for the purpose of optimizing the performance and energy efficiency of machine learning inference at the edge. The cloud exploits a novel multimodal matching network model that outperforms six state-of-the-art stress recognition algorithms by 2%–7% in terms of accuracy. An offloading decision process is formulated to strike the right balance between accuracy, latency, and energy. By addressing the interplay of edge–cloud, the proposed hierarchical solution leads to a reduction of 77.89% in response time and 78.56% in energy consumption with only a 7.6% drop in accuracy compared to the Internet of Things (IoT)–Cloud scheme, and it achieves a 5.8% increase in accuracy on average compared to the IoT-Edge scheme.