Privacy-preserving Method Research Articles

Effective Route Recommendation Leveraging Differentially Private Location Data

The proliferation of GPS-enabled devices and advances in positioning technologies have greatly facilitated the collection of user location data, making them valuable across various domains. One of the most common and practical uses of these location datasets is to recommend the most probable route between two locations to users. Traditional algorithms for route recommendation rely on true trajectory data collected from users, which raises significant privacy concerns due to the personal information often contained in location data. Therefore, in this paper, we propose a novel framework for computing optimal routes using location data collected through differential privacy (DP)-based privacy-preserving methods. The proposed framework introduces a method for accurately extracting transitional probabilities from perturbed trajectory datasets, addressing the challenge of low data utility caused by DP-based methods. Specifically, to effectively compute transitional probabilities, we present a density-adjusted sampling method that enables the collection of representative data across all areas. In addition, we introduce an effective scheme to approximately estimate transitional probabilities based on sampled datasets. Experimental results on real-world data demonstrate the practical applicability and effectiveness of our framework in computing optimal routes while preserving user privacy.

Daymet and DeGAUSS: Development of a Novel Method that Links High Spatial Resolution Gridded Meteorological Data with Geocoded Healthcare Encounters

ObjectiveDue to human-caused climate change, extreme temperature-related morbidity/mortality are increasing global public health concerns. To better understand the burdens of heat exposure, researchers must merge accurate/granular environmental data with clinical data. We developed a privacy-preserving method that links high-spatial-resolution continuous, gridded, meteorological data to geocoded healthcare encounters. ApproachIn an R environment, we developed code that links geocoded healthcare encounters to publicly-available Daymet data via the corresponding Daymet raster cell number and healthcare encounter date. Daymet provides daily 1km x 1km surface meteorological data over North America since 1980. User specifications include the bounding box latitude/longitude of Daymet data download, which Daymet environmental variables are linked, and the day-lag of Daymet data matching to healthcare encounter date. ResultsLinking 500,000 healthcare encounters over a 7-day lag with 7 years of Daymet data comprising minimum/maximum temperature over a bounding box encompassing Cook County, Illinois (5,670 1km raster cells) took 11.92 minutes with 4 CPUs and 8GB RAM. Thirteen years of Daymet data took 23.85 minutes. Linking 7.5 million encounters with 7 years of data took 28.49 minutes. ConclusionsOur newly-developed method for linking Daymet data to geocoded healthcare encounters preserves patient privacy, and processes the data efficiently over a minimal timespan. ImplicationsDaymet data linked to healthcare encounters allows for accurate assessment of the association between environmental variable exposure and any clinical outcome over North America. Packaging the code allows for public distribution within the Decentralized Geomarker Assessment for Multi-Site Studies (DeGAUSS) universe of geoanalytic tools.

Trajectory-aware privacy-preserving method with local differential privacy in crowdsourcing

In spatial crowdsourcing services, the trajectories of the workers are sent to a central server to provide more personalized services. However, for the honest-but-curious servers, it also poses a challenge in terms of potential privacy leakage of the workers. Local differential privacy (LDP) is currently the latest technique to protect data privacy. However, most of LDP-based schemes have limitations in providing good utility due to extensive noise in perturbing trajectories. In this work, to balance the privacy and utility, we propose a novel pattern-aware privacy protection method called trajectory-aware privacy-preserving with local differential privacy (TALDP). The key idea is that, rather than applying the same degree of perturbation to all location points, we employ adaptive privacy budget allocation, assigning varied privacy budgets to individual location points, thereby mitigating the perturbation’s impact and enhancing overall utility. Meanwhile, to ensure the privacy, we give the different perturbing points to different privacy budgets according to their important degree for the patterns of the trajectories. In particular, we use Karman filter method to select the important location points and decide their privacy budgets. We conduct extensive experiments on three real datasets. The results show that our approach improves the utility over many other current methods while still provide good the privacy protection.

Credit risk: A new privacy-preserving decentralized credit assessment model

To facilitate the collaborative modelling of privacy-preserving credit assessments under multi-party data sharing, this study suggests a privacy-preserving method for the collaborative modelling of linear regression credit assessments that is based on random invertible matrix transformation. The results indicate that the proposed method is capable of effectively protecting the original data information from being compromised during the collaborative modelling process. Besides, our model can integrate multidimensional data to alter the data distribution by employing random invertible matrix transformation. Finally, our collaborative modelling method can incorporate multi-party data to train a linear regression credit assessment model while maintaining data privacy.

Privacy preservation network with global-aware focal loss for Interactive Personal Visual Privacy Preservation

Existing visual privacy preservation methods often encounter the challenge of over-protection, which usually fails to accurately protect a specified person in an image. This paper introduces a novel task called “Interactive Personal Visual Privacy Preservation (IPVPP)”, with the objective of safeguarding a desired individual in an image based on user-click prompts. We propose a new framework, Privacy Preservation Network (PPNet) with Global-aware Focal Loss, tailored for IPVPP task. To address the sample imbalance issue in IPVPP, we present a Global-aware Focal Loss (GFL), which introduces global and local difficulty balance weights to dynamically adjust the overall decision boundary, thereby mitigating the sample imbalance problem for better optimization of models. Furthermore, to tackle the insufficient fusion between click and visual features, we propose a Click-Image Fusion (CIF) module, which effectively emphasizes the interaction between click and image features, providing robust query inputs to the model. We finally employ generative adversarial networks to ensure the reusability of privacy images, striking a balance between privacy preservation and image value. Experimental results on the DAVIS dataset demonstrate the effectiveness of PPNet, achieving improvements measured in NoC 90, IS, SSIM, F-1, and DIR, with 0.22, 0.19, 0.009, 4.23, and 2.15 improvements, respectively.

Gradient Inversion Attacks: Impact Factors Analyses and Privacy Enhancement.

Gradient inversion attacks (GIAs) have posed significant challenges to the emerging paradigm of distributed learning, which aims to reconstruct the private training data of clients (participating parties in distributed training) through the shared parameters. For counteracting GIAs, a large number of privacy-preserving methods for distributed learning scenario have emerged. However, these methods have significant limitations, either compromising the usability of global model or consuming substantial additional computational resources. Furthermore, despite the extensive efforts dedicated to defense methods, the underlying causes of data leakage in distributed learning still have not been thoroughly investigated. Therefore, this paper tries to reveal the potential reasons behind the successful implementation of existing GIAs, explore variations in the robustness of models against GIAs during the training process, and investigate the impact of different model structures on attack performance. After these explorations and analyses, this paper propose a plug-and-play GIAs defense method, which augments the training data by a designed vicinal distribution. Sufficient empirical experiments demonstrate that this easy-toimplement method can ensure the basic level of privacy without compromising the usability of global model.

Research on Privacy Issues in Smart Metering System: An Improved TCN-Based NILM Attack Method and Practical DRL-Based Rechargeable Battery Assisted Privacy Preserving Method

Smart meters, as a key component of Advanced Metering Infrastructure (AMI), collect fine-grained electricity consumption data for demand response in smart grids. While this data improves the grid’s accuracy, it also poses significant threats to users’ privacy. In this paper, we study the privacy issue in smart metering systems from both attacker and defender perspectives. First, we propose an improved Temporal Convolutional Network (TCN) based Non-Intrusive Load Monitoring (NILM) attack method, which infers electrical appliance usage from public load curves, addressing the gradient vanishing, gradient exploding, and other problems while improving attack accuracy. Second, we develop a rechargeable battery-assisted energy management system to hide load characteristics of electrical appliances by adding physical noise, thus resisting NILM attacks. To address the privacy-cost trade-off optimization problem, we propose a Practical Deep Reinforcement Learning-based Rechargeable Battery assist Privacy Preserving Method (PRoP) that learns optimal battery charging/discharging policies. We design a novel privacy measurement method and constraints to ensure the feasibility of system deployment and prove PRoP’s effectiveness in resisting NILM attacks. Comprehensive evaluations demonstrate that our improved TCN-based NILM method achieves an attack success rate of over 80% on various electrical appliances, improving attack performance (MAE, RMSE) by 20% compared to existing methods while reducing model training time. Moreover, our proposed PRoP achieves a better trade-off between privacy protection and electricity cost than existing battery-assisted methods, reducing costs by 5% and the attack success ratio to 36%, while increasing the MAE and RMAE obtained by NILM by 3 times. <italic xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">Note to Practitioners</i> —Smart grid, which can support bidirectional information transmission, has a series of advantages, such as high efficiency and high stability. However, it also brings a significant threat to users’ electricity privacy. Although encryption-based privacy protection methods have been deployed on terminal devices of smart grids to prevent privacy leaks, this method can often only defend against intrusive attacks and has little effect on non-intrusive attacks. To this end, this paper studies the privacy issues caused by non-intrusive attacks. Specifically, to better study the protection method, we first investigate the attack mechanism and design an improved TCN-based Non-Intrusive Load Monitoring method. Then, we propose a Practical Reinforcement learning-based rechargeable battery-assisted Privacy preserving method (PRoP) to defend against this attack physically. The most practical contribution of this paper is that, compared with existing battery-assisted privacy protection methods, we do not blindly pursue algorithm performance but fully consider the practical factors of deployment, such as limiting battery capacity and constraining battery charging and discharging behavior. This can guide practitioners to better apply this technology in practice.

Privacy-preserving method for sensitive partitions of electricity consumption data based on hybrid differential privacy and k-anonymity

Abstract With the development of smart grid, traditional privacy protection methods are no longer sufficient to cope with the mining and analysing of users’ electricity consumption data, which can easily lead to the leakage of users’ privacy. In this paper, for the challenge of privacy protection of user electricity data in smart grid, we propose a privacy protection method for sensitive partition of electricity data based on hybrid differential privacy and k-anonymity technique, which identifies pattern-specific quasi-identifiers by random forest technique and divides the data into privacy-invasive and non-privacy-invasive partitions, and adopts the Laplacian noise addition (numerical quasi-identifiers) and data generalisation (sub-types of quasi-identifiers) techniques to achieve differential privacy protection for different partitions. Finally, it is experimentally verified that the method performs better in terms of information reduction accuracy and execution time, effectively striking a balance between protecting user privacy and maintaining data utility.

Privacy-preserving State of Health prediction for electric vehicle batteries: A comprehensive review

This work presents an in-depth systematic literature review of privacy-preserving machine learning techniques for battery State of Health (SoH) prediction in electric vehicles (EVs). Our review methodically analyses a corpus of 43 research articles and publications spanning from 2012 to 2024 which provides a comprehensive landscape of the field’s evolution and current trends. This review covers a range of machine learning techniques and uniquely addresses first-of-its-kind privacy needs specific to SoH prediction for EVs, with a detailed examination of the integration of privacy-preserving methods. We also delve into the key challenges associated with data processing and feature engineering, and how these challenges are managed. The methodology employed includes not only the identification but also a thematic analysis of the literature, focusing on how these techniques balance data confidentiality with analytical performance. We proposed a privacy preserving system model for SoH prediction. The review is further enriched by various real-time applications and case studies, illustrating the practical implementation and efficacy of the discussed methods. This systematic approach provides a well-rounded view of the existing solutions and highlights gaps for future research, offering invaluable insights for both academic and industrial stakeholders in the domain of electric vehicle technology.

Responsible AI for cardiovascular disease detection: Towards a privacy-preserving and interpretable model

Background and objectiveCardiovascular disease (CD) is a major global health concern, affecting millions with symptoms like fatigue and chest discomfort. Timely identification is crucial due to its significant contribution to global mortality. In healthcare, artificial intelligence (AI) holds promise for advancing disease risk assessment and treatment outcome prediction. However, machine learning (ML) evolution raises concerns about data privacy and biases, especially in sensitive healthcare applications. The objective is to develop and implement a responsible AI model for CD prediction that prioritize patient privacy, security, ensuring transparency, explainability, fairness, and ethical adherence in healthcare applications. MethodsTo predict CD while prioritizing patient privacy, our study employed data anonymization involved adding Laplace noise to sensitive features like age and gender. The anonymized dataset underwent analysis using a differential privacy (DP) framework to preserve data privacy. DP ensured confidentiality while extracting insights. Compared with Logistic Regression (LR), Gaussian Naïve Bayes (GNB), and Random Forest (RF), the methodology integrated feature selection, statistical analysis, and SHapley Additive exPlanations (SHAP) and Local Interpretable Model-agnostic Explanations (LIME) for interpretability. This approach facilitates transparent and interpretable AI decision-making, aligning with responsible AI development principles. Overall, it combines privacy preservation, interpretability, and ethical considerations for accurate CD predictions. ResultsOur investigations from the DP framework with LR were promising, with an area under curve (AUC) of 0.848 ± 0.03, an accuracy of 0.797 ± 0.02, precision at 0.789 ± 0.02, recall at 0.797 ± 0.02, and an F1 score of 0.787 ± 0.02, with a comparable performance with the non-privacy framework. The SHAP and LIME based results support clinical findings, show a commitment to transparent and interpretable AI decision-making, and aligns with the principles of responsible AI development. ConclusionsOur study endorses a novel approach in predicting CD, amalgamating data anonymization, privacy-preserving methods, interpretability tools SHAP, LIME, and ethical considerations. This responsible AI framework ensures accurate predictions, privacy preservation, and user trust, underscoring the significance of comprehensive and transparent ML models in healthcare. Therefore, this research empowers the ability to forecast CD, providing a vital lifeline to millions of CD patients globally and potentially preventing numerous fatalities.

Improvement of data analysis and protection using novel privacy-preserving methods for big data application

Due to the increasing volume of data, the importance of data analysis systems has become more critical. An intrusion detection system is a type of software that monitors and analyzes the data collected by a network or system. Due to the increasing volume of data collected in the medical field, it has become harder for traditional methods to detect unauthorized access and manipulation of the data. To advance the efficiency of big data analysis, various techniques are used in IDS. This paper proposes a method that combines the deep learning network and proposed optimization algorithm. The goal of this paper is to develop a classification model that takes into account the hidden layer nodes of the DBN and then implement a PSO algorithm to improve its structure. The results of the simulations show that the Spark-DBN-PSO algorithm achieves a 99.04% accuracy rate, which is higher than the accuracy of other deep neural network (DNN) and artificial neural network (ANN) algorithms. The results of the research demonstrate that the proposed methodology performs superior than the existing algorithm.

A Novel Privacy Preservation Scheme by Matrix Factorized Deep Autoencoder

Data transport entails substantial security to avoid unauthorized snooping as data mining yields important and quite often sensitive information that must be and can be secured using one of the myriad Data Privacy Preservation methods. This study aspires to provide new knowledge to the study of protecting personal information. The key contributions of the work are an imputation method for filling in missing data before learning item profiles and the optimization of the Deep Auto-encoded NMF with a customizable learning rate. We used Bayesian inference to assess imputation for data with 13%, 26%, and 52% missing at random. By correcting any inherent biases, the results of decomposition problems may be enhanced. As the statistical analysis tool, MAPE is used. The proposed approach is evaluated on the Wiki dataset and the traffic dataset, against state-of-the-art techniques including BATF, BGCP, BCPF, and modified PARAFAC, all of which use a Bayesian Gaussian tensor factorization. Using this approach, the MAPE index is decreased for data which avails privacy safeguards than its corresponding original forms.

ZKAV:Zero Knowledge proof for AV

This study investigates how blockchain technology can be used in the automotive industry and society, with a focus on addressing privacy concerns and compliance issues. The study investigates the application of privacy-preserving methods, including zero-knowledge proofs and anonymous credentials, to protect vehicle data shared with servers and prevent V2X channel spam. The proposed blockchain ecosystem involves collaborations among stakeholders like governmental bodies and automotive industry participants to ensure user privacy, implementation employs Ethereum[5], circom, and ezkl for practical realization, the study demonstrates the potential of blockchain and [7] Zero Knowledge Proofs in overcoming challenges, promoting a more secure digital environment.

Privacy-preserving sports data fusion and prediction with smart devices in distributed environment

With the rapid advancement of sports analytics and fan engagement technologies, the volume and diversity of physique data generated by smart devices across various distributed sports platforms have grown significantly. Extracting insights and enhancing fan experiences from such data offer considerable benefits. Yet, this process unveils two primary challenges. Firstly, efficiently utilizing the vast datasets in sports analytics is daunting due to the complex nature of the sports industry. Secondly, the data collected from diverse sources and stored in distributed platforms contain sensitive information like fan preferences and athlete performance metrics, posing risks of privacy breaches. To address these challenges, we leverage an advanced Locality-Sensitive Hashing technique, known as PSDFPALSH\\documentclass[12pt]{minimal} \\usepackage{amsmath} \\usepackage{wasysym} \\usepackage{amsfonts} \\usepackage{amssymb} \\usepackage{amsbsy} \\usepackage{mathrsfs} \\usepackage{upgreek} \\setlength{\\oddsidemargin}{-69pt} \\begin{document}$$_{\ ext {ALSH}}$$\\end{document}, tailored for the sports domain. This paper presents a new privacy-preserving method for sports data fusion and prediction in distributed environments, utilizing enhanced Locality-Sensitive Hashing to protect sensitive information while maintaining high data utility. Through extensive experimentation, our approach demonstrates superior performance over existing methods in terms of Mean Absolute Error (MAE), Root Mean Squared Error (RMSE), and computational efficiency.

Building Communication Efficient Asynchronous Peer-to-Peer Federated LLMs with Blockchain

Large language models (LLM) have gathered attention with the advent of ChatGPT. However, developing personalized LLM models faces challenges in real-world applications due to data scarcity and privacy concerns. Federated learning addresses these issues, providing collaborative training while preserving the client’s data. Although it has made significant progress, federated learning still faces ongoing challenges, such as communication efficiency, heterogeneous data, and privacy-preserving methods. This paper presents a novel, fully decentralized federated learning framework for LLMs to address these challenges. We utilize different blockchain-federated LLM (BC-FL) algorithms, effectively balancing the trade-off between latency and accuracy in a decentralized-federated learning environment. Additionally, we address the challenge of communication overhead in peer-to-peer networks by optimizing the path for weight transfer and mitigating node anomalies. We conducted experiments to evaluate memory usage and latency in server and serverless environments. Our results demonstrate a decrease in latency by 5X and a 13% increase in accuracy for serverless cases. Comparisons between synchronous and asynchronous scenarios revealed a 76% reduction in information passing time for the latter. The PageRank method is most efficient in eliminating anomalous nodes for better performance of the global federated LLM model. The code is available on GitHub (https://github.com/Sreebhargavibalijaa/Federated_finetuning_LLM-s_p2p_environment)

Privacy-Preserving Breast Cancer Prediction Based on Logistic Regression

Abstract With the increasing strain on today’s healthcare resources, there is a growing demand for pre-diagnosis testing. In response, researchers have suggested diverse machine learning models for disease prediction, among which logistic regression stands out as one of the most effective models. Its objective is to enhance the accuracy and efficiency of pre-diagnosis testing, thereby alleviating the burden on healthcare resources. However, when multiple medical institutions collaborate to train models, the untrusted cloud server may pose a risk of private data leakage, enabling participants to steal data from one another. Existing privacy-preserving methods often suffer from drawbacks such as high communication costs, long training times and lack of security proofs. Therefore, it is imperative to jointly train an excellent model collaboratively and uphold data privacy. In this paper, we develop a highly optimized two-party logistic regression algorithm based on CKKS scheme. The algorithm optimizes ciphertext operations by employing ciphertext segmentation and minimizing the multiplication depth, resulting in time savings. Furthermore, it utilizes least squares to approximate sigmoid functions within specific intervals that cannot be handled by homomorphic encryption. Finally, the proposed algorithm is evaluated on a breast cancer dataset, and simulation experiments demonstrate that the model’s prediction accuracy, after machine learning training, exceeds 96% for two-sided encrypted data.

Data Security and Privacy in Genomics Research: A Comparative Analysis to Protect Confidentiality

The quick progress of genomics examination has driven a surge in the creation of significantly fragile genomic data, making ensuring its security essential. This data contains sensitive information roughly an individual's prosperity, family history, and defencelessness to ailments. Unauthorized access or mishandling can lead to isolation, stigmatization, and mystery breaches. The potential threats to genomic data affirmation are multifaceted, checking the chance of re-identification and extended defense lessness to data breaches, hacking events, and unauthorized get to by harmful actors. To address these challenges, a multifaceted approach is required, tallying solid privacy-preserving methods, securing data capacity, and transmission sharpens, and getting to controls. Encryption techniques, differential security methods, and secure multiparty computation offer promising streets for securing genomic data while progressing collaborative ask approximately. Establishing clear authority frameworks and rules for data management, capacity, and sharing is essential to reduce security threats in genomics research. Collaboration between researchers, policymakers, industry partners, and support groups is essential for developing comprehensive methods to protect genomic data security. By prioritizing security concerns and executing effective safeguards, the community can uphold individuals' rights, maintain open acceptance, and drive advancements in genomics research for the betterment of society.

Deep learning-based fall detection using commodity Wi-Fi

As the phenomenon of an aging population gradually becomes common worldwide, the pressure on the elderly has seen a notable increase. To address this challenge, fall detection systems are important in ensuring the safety of the elderly population, particularly those living alone. Wi-Fi sensing, as a privacy-preserving method of perception, can be deployed indoors for detecting human activities such as falls, based on the reflective properties of electromagnetic waves. Signals generated by transmitters experience reflections from various objects within indoor environments, leading to distinct propagation paths. These signals eventually aggregate at the receiver, incorporating details about the objects’ orientation and their activity states. In this study, within practical experimental environments, we collect dataset and utilize deep learning method to classify the falling events.

An enhanced discrete particle swarm optimization for structural k-Anonymity in social networks

Recently, social network usage has exhibited explosive growth, leading to a huge amount of users’ private data available. The main challenge in releasing social network data publicly is the protection of the users’ privacy while preserving its utility for third parties. Accordingly, several social network privacy-preserving methods have been introduced, where anonymization is the most common approach. Structural k-anonymity is a widely used anonymization model to mask the structure of social networks by clustering the edges and nodes into super-edges and super-nodes. However, it comes at the cost of losing structural information, which is measured by a criterion called structural information loss (SIL). This study introduces an enhanced discrete particle swarm optimization (EDPSO) algorithm, which effectively minimizes the SIL within the clustering process of the structural k-anonymity model, leading to a high-utility anonymized network. In this regard, we propose a vector-based solution representation that can be efficiently exploited by the EDPSO. Moreover, a novel position updating heuristic is suggested for the EDPSO, which adaptively tunes the operators’ selection probabilities. This happens based on each operator’s performance both in the current iteration and their history regarding the number and the average amount of fitness improvements in the previous iterations. We also propose two fortified versions of the EDPSO algorithm (EDPSOVNS and EDPSOSA) by employing two new network-specific local search strategies to enhance the exploration, exploitation, and convergence rate of the process. Simulation results on the nine real-world networks demonstrate the superiority of the suggested algorithms in terms of the fitness value, reliability, and convergence rate over other analyzed approaches found in the literature.

ESVFL: Efficient and secure verifiable federated learning with privacy-preserving

Federated learning has been widely applied as a distributed machine learning method in various fields, allowing a global model to be trained by sharing local gradients instead of raw data. However, direct sharing of local gradients still carries the risk of privacy data leakage, and the malicious server might falsify aggregated result to disrupt model updates. To address these issues, a lot of privacy-preserving and verifiable federated learning schemes have been proposed. However, existing schemes suffer from significant computation overhead in either encryption or verification. In this paper, we present ESVFL, an efficient and secure verifiable federated learning scheme with privacy-preserving. This scheme can simultaneously achieve low computation overhead for encryption and verification on the user side. We design an efficient privacy-preserving method to encrypt the users’ local gradients. Using this method, the computation and communication overheads of encryption on the user side is independent of the number of users. Users can efficiently verify the correctness of aggregated results returned by the cloud servers using cross-verification. During the verification process, there is no interaction among users and no additional computation is required. Furthermore, we also construct an efficient method to address the issue of user dropout. When some users drop out, online users do not incur any additional computation and communication overheads, while guaranteeing the correctness of the aggregated result of online users’ encrypted gradients. The security analysis and the performance evaluation demonstrate that ESVFL is secure and can achieve efficient encryption and verification.