Federated learning data protection scheme based on personalized differential privacy in psychological evaluation

BackgroundHealth physical examinations play a crucial role in early detection of cancer and chronic disease. However, privacy concerns limit the utilization of this kind of data for health interventions and research. Synthetic data methods based on differential privacy are increasingly used to create complete datasets that protect privacy while enabling data analysis and result interpretation. Hence, the use of synthetic algorithms based on differential privacy for privacy protection of physical examination data is a promising research direction.MethodsThree synthetic algorithms, PrivBayes, PeGS, and DP-Gibbs were used to generate complete synthetic datasets that adhere to differential privacy standards using physical examination data composed of categorical data, which compared with the existing algorithm Private-PGM.ResultsCompared with the existing algorithm, DP-Gibbs can provide privacy preserving capacity of 4.686 (ε = 0.5), while the existing algorithm only with 2.012. In addition, DP-Gibbs provides 0.620 of precision, 0.539 of F1-score, 0.342 of Kappa Coefficient, and 0.765 of AUC-score. The corresponding statistical results of existing algorithm are 0.520, 0.321, 0.188 and 0.695.ConclusionsThe main contributions of this study are the exploration of combination models incorporating different noise forms and Bayesian synthetic algorithms, alongside a comparative analysis against existing algorithms. This study explored the balance between privacy protection and data utility under different levels of privacy protection, and DP-Gibbs offers more stable technical support for de-identifying physical examination data prior to sharing and analysis, which realized the mining and application of a wider range of medical data under the requirements of privacy protection. By leveraging this effective privacy protection technique, clinical researchers can extract valuable insights on diseases and population health from the physical examination data without the risk of leaking private information.Supplementary InformationThe online version contains supplementary material available at 10.1186/s12911-025-03109-1.

У статті здійснено когнітивне моделювання розвитку загроз приватності особистості та безпеці країни. Для моделювання обрано такі фактори: «Приватність особистості», «Безпека розвитку країни», «Розробка систем захисту», «Онлайн-присутність», «Кібер-атаки», «Втрати від кіберзлочинів», «Витрати на захист», «Сталий розвиток». Результати показали, що всі фактори в системі є дестабілізаційними й чутливими до змін. Найбільш чутливим є фактор «Онлайн-присутність». Такий характер системи свідчить про те, що будь-які зміни можуть дестабілізувати систему, привести до змін усі інші фактори. При цьому система сама сприятиме таким змінам, оскільки є нестійкою. Отже, уся система забезпечення захисту приватності особистості та безпеки країни потребує постійної уваги й моніторингу змін усіх факторів. Фактори «Приватність особистості» та «Безпека розвитку країни» є взаємозалежними, оскільки демонструють майже однаковий вплив на них усіх інших факторів системи. Так, ці фактори підсилюватимуться при зростанні факторів «Розробка систем захисту», «Онлайн-присутність», «Витрати на захист» і «Сталий розвиток». Зростання рівня захисту приватності особистості сприятиме зростанню рівня безпеки країни й навпаки. При цьому «Приватність особистості» сильніше впливає на зміцнення рівня безпеки країни, порівняно з впливом безпеки держави на захист приватності особистості при її присутності в мережі Інтернет. Лише два фактори – «Кібератаки» та «Втрати від кіберзлочинів» зменшуватимуть рівень захисту приватностій безпеки країни при їх зростанні. Рівень захисту приватності є більш чутливим до впливу на неї з боку кібератак, порівняно з чутливістю рівня безпеки країни. На рівень безпеки розвитку держави найбільше впливає рівень захисту приватності особистості, і цей вплив на 14 % більш потужний, порівняно з протилежним впливом факторів. Аналіз засвідчив, що основним пріоритетом під час розвитку мережевих технологій, упровадженні ІКТ у всі сфери суспільного життя та розвитку інформаційної економіки повинно бути забезпечення захисту приватності особистості, її персональних даних. Забезпечення безпеки розвитку країни потребуватиме більшого рівня розвитку систем захисту, порівняно із захистом приватності особистості, при цьому на 4 % менше впливатиме на сталий розвиток.

In response to the privacy leakage risks inherent in the big data processing of power user personas, propose a collaborative optimization‐based joint privacy protection mechanism for Differential Privacy (DP) and Secret Sharing (SS). This method reveals the inherent correlation between DP noise parameters and SS sharding strategy and designs an adaptive noise sharding mapping mechanism to achieve joint optimization of privacy protection and communication overhead. Based on this mechanism, a specific privacy protection scheme for power user personas is designed. Experimental results demonstrate that the DP + SS privacy protection method designed in this study achieves an accuracy rate of 93.14% on the MNIST dataset and maintains an accuracy rate of 92.06% across multiple datasets, indicating that the scheme effectively quantifies the level of privacy protection. The privacy protection level of the scheme is quantitatively defined through differential privacy budget ε ∈ [0.11,0.15], secret sharing ( k = 3, n = 5) threshold mechanism, and privacy leakage probability δ = 10 −5 . The privacy budget is dynamically allocated through the correlation between differential privacy noise variance and the minimum reconstruction threshold of secret sharing, ensuring that privacy risks are controllable and quantifiable. At the same time, the average training time for servers is about 186.4 s, which refers to the total end‐to‐end training time in a four‐server cluster environment. Compared with a single server encryption scheme, the efficiency is improved by 48.6%. The communication overhead has increased to 10.6GB, which is the total chip transmission overhead per 10,000 users. However, the calculation stability can be maintained through load balancing, indicating that the scheme has certain feasibility. The aforementioned research results not only provide a viable applied cryptographic technology solution for privacy protection in power big data but also offer new insights into balancing privacy protection strength and data availability. More importantly, the joint privacy protection mechanism proposed by the research institute can serve as a fundamental enabling technology for the infrastructure of a secure smart grid, providing quantifiable and trustworthy privacy and security guarantees for advanced data‐driven applications such as precise load prediction, demand‐side response, and collaborative optimization of distributed energy, thereby supporting the reliable and intelligent development of the smart grid.

Federated learning (FL), which enables multiple distributed devices (clients) to collaboratively train a global model without transmitting their private data, has attracted much attention in the Internet of Things (IoT) domain. Compared with centralized learning, FL has obvious privacy advantages because it can protect the clients’ raw data from direct access by adversaries. Furthermore, to prevent the adversaries from inferring private information from the transmitted parameters, several FL algorithms based on differential privacy (DP) have been proposed, where the clients add artificial noise to their local parameters for privacy protection. Nevertheless, the added noise would disrupt the learning process and degrade the performance of the trained model. Considering this, in this article, we develop a performance-enhanced DP-based FL (PEDPFL) algorithm, where a classifier-perturbation regularization method is proposed to improve the robustness of the trained model against DP-injected noise. We derive the theoretical privacy and convergence analysis of the proposed algorithm, and also demonstrate the influence of some hyperparameters on the convergence performance. Simulation results on real-world data sets show that the proposed algorithm has better classification performance than the existing DP-based FL algorithms at the same level of privacy protection, and thus, it is more applicable to IoT applications.

An experimental tattoo de-identification system for privacy protection in still images is described in the paper. The system consists of the following modules: skin detection, region of interest detection, feature extraction, tattoo database, matching, tattoo detection, skin swapping, and quality evaluation. Two methods for tattoo localization are presented. The first is a simple ad-hoc method based only on skin colour. The second is based on skin colour, texture and SIFT features. The appearance of each tattoo area is de-identified in such a way that its skin colour and skin texture are similar to the surrounding skin area. Experimental results for still images in which tattoo location, distance, size, illumination, and motion blur have large variability are presented. The system is subjectively evaluated based on the results of tattoo localization, the level of privacy protection and the naturalness of the de-identified still images. The level of privacy protection is estimated based on the quality of the removal of the tattoo appearance and the concealment of its location.

Probabilistic matrix factorization with personalized differential privacy

Differential privacy is often used in location privacy protection because of its strict reasoning and proof privacy guarantee. When users make continuous location query, it will cause noise superposition, which leads to the decline of query accuracy. At present, although differential privacy based on rule tree structure can reduce the query error, it will generate a lot of invalid zero nodes. The data structure is too large, and more improvement in the query accuracy can be further investigated. In this paper, we proposed a differential privacy location privacy protection method based on generative adversary network. Firstly, the definition of location data privacy protection under differential privacy mechanism is given, and then resume density aware network under differential privacy mechanism. Based on density aware network, the privacy protection problem of location data can be transformed into the distribution of fitting trajectory length. Finally, we use Markov chain to generate a new trajectory, and introduce the generative adversarial network to construct a set differential privacy protection method. Compared with other methods to improve the accuracy of differential privacy query, this method can effectively reduce the problem of query accuracy decline caused by noise superposition in continuous query, and can adapt to lbs location query service in different density environments.KeywordsGenerative adversarial networkLocation differencePrivacy protectionMarkov chainDensity aware network

Privacy protection in video images is becoming one of the research focuses in the field of remote collaborative system. In this paper, a method for face detection, tracking and privacy protection is presented. According to skin-color distribution in the color space, we developed a statistical skin-color model through interactive sample training. Using this model we convert the color image to binary image and then segment face candidate region. Then we use a facial feature matching scheme for further detection. The presence or absence of a face in each region is verified by means of mouth detector. Real time detection and tracking can be achieved by using this method in video images. In order to speed up tracking, we improve the traditional method by adding motion prediction, which works better when several disturbing objects appear simultaneously. Finally we make the tracking region blurring and transmit the frames to the remote collaborative sites to obtain the privacy protection. The level of privacy protection can be dynamically adjusted according to collaborators' requests and credibility of remote sites. The experiment results show the proposed method not only has high speed and efficiency, but also is robust to head rotation to some extent.

The essential issues of data silos and user privacy leakage could be relaxed substantially by the development of the federated learning (FL) architecture. In a collaborative multi-user modeling situation, malicious attackers could still use user gradient information to infer the danger of user privacy. To mitigate the issue of privacy leakage, differential privacy (DP) mechanism is integrated into the federated learning framework to assess privacy loss and introduce noise to the local model parameters of users. In addition, in order to minimize information leakage and provide better noise rejection, Rényi differential privacy (RDP) is introduced as a privacy metric, which improves the balance between model privacy and utility. Owing to the unknown target model and limited communication cost resources, a client-based adaptive learning algorithm is developed in which each local model parameter is adaptively updated during local iterations to accelerate model convergence and avoid model overfitting. The experimental results reveal that the client-based adaptive federation learning model in this paper outperforms the classic model at a fixed communication cost, is more robust to noise resistance and variable hyperparameter settings, and provides more accurate privacy protection during transmission.

The popularity of wearable devices equipped with a variety of sensors that can measure users’ health status and monitor their lifestyle has been increasing. In fact, healthcare service providers have been utilizing these devices as a primary means to collect considerable health data from users. Although the health data collected via wearable devices are useful for providing healthcare services, the indiscriminate collection of an individual’s health data raises serious privacy concerns. This is because the health data measured and monitored by wearable devices contain sensitive information related to the wearer’s personal health and lifestyle. Therefore, we propose a method to aggregate health data obtained from users’ wearable devices in a privacy-preserving manner. The proposed method leverages local differential privacy, which is a de facto standard for privacy-preserving data processing and aggregation, to collect sensitive health data. In particular, to mitigate the error incurred by the perturbation mechanism of location differential privacy, the proposed scheme first samples a small number of salient data that best represents the original health data, after which the scheme collects the sampled salient data instead of the entire set of health data. Our experimental results show that the proposed sampling-based collection scheme achieves significant improvement in the estimated accuracy when compared with straightforward solutions. Furthermore, the experimental results verify that an effective tradeoff between the level of privacy protection and the accuracy of aggregate statistics can be achieved with the proposed approach.

With the continuous development of edge computing, the application scope of mobile crowdsourcing (MCS) is constantly increasing. The distributed nature of edge computing can transmit data at the edge of processing to meet the needs of low latency. The trustworthiness of the third-party platform will affect the level of privacy protection, because managers of the platform may disclose the information of workers. Anonymous servers also belong to third-party platforms. For unreal third-party platforms, this paper recommends that workers first use the localized differential privacy mechanism to interfere with the real location information, and then upload it to an anonymous server to request services, called the localized differential anonymous privacy protection mechanism (LDNP). The two privacy protection mechanisms further enhance privacy protection, but exacerbate the loss of service quality. Therefore, this paper proposes to give corresponding compensation based on the authenticity of the location information uploaded by workers, so as to encourage more workers to upload real location information. Through comparative experiments on real data, the LDNP algorithm not only protects the location privacy of workers, but also maintains the availability of data. The simulation experiment verifies the effectiveness of the incentive mechanism.

The existing privacy protection schemes for Location‐Based Service (LBS) only protect users' location privacy or query privacy, which can not adopt both of the privacy protections simultaneously in the LBS system. Moreover, these schemes cannot take into account the spatial‐temporal correlation and background knowledge. In response to the above mentioned questions, the LBS Privacy Protection Scheme Based on Differential Privacy (DPLQ) is proposed. The method contains two kinds of privacy protection algorithms: users' location privacy protection algorithm and users' query privacy protection algorithm. The users' location privacy protection algorithm divides the map using the Voronoi diagram, choosing l fake location points based on the improved k ‐means algorithm and l ‐diversity idea, and protects users' location privacy with the Laplace mechanism. Based on the k ‐anonymous algorithm, the users' query privacy protection algorithm builds a query k ‐anonymous set according to the neighbour users' query requests at the same time t in the cluster and the historical query probability of the region’s POI and protects users' query privacy with the exponential mechanism. Through setting the privacy protection intensity of the algorithm by the users, the generated location dataset and query k ‐anonymous set can resist a variety of attacks from malicious attackers. Theoretical analysis and experimental results show that the scheme can effectively protect the location privacy and query privacy of users.

The current privacy protection methods adopted by social network providers rely on restricting users' access rights. They force a user to rigidly divide other users into two categories only: friends and strangers. Based on this classification, they prevent non-friends from accessing to the user's data while provide full access for friends regardless of how close they are to her. However, the level of privacy protection can be increased gradually and smoothly rather than firmly like a zero/one function. Moreover, the utility of social networks is reduced if we prevent data miners from computing global statistics by applying rigid privacy policies. In this paper, we present a distance-based location privacy protection system (DBLP2) to preserve location privacy of social network users based on their friendship distance. Whenever a user wants to see another user's location (in her profile or from spatiotemporal tags on her posts), the system returns a differentially private response based on their friendship distance. In our proposed system, a user's location information provided for other users is more generalized as the friendship distance increases. In other words, family members and close friends receive a more accurate response than casual friends and strangers. Through analysis, we show that our proposed system makes the process of location privacy protection more flexible in terms of friendship distances.

With the mass deployment of cameras, concern has risen about protecting a person's privacy as he goes about his daily life. Many of the cameras are installed to perform surveillance tasks that do not require the identity of a person. In the context of surveillance applications, we examine the trade-off between privacy and security. The trade-off is accomplished by looking at quantitative measures of privacy and surveillance performance. To provide privacy protection we examine the effect on surveillance performance of a parametric family of privacy function. A privacy function degrades images to make identification more difficult. By varying the parameter, different levels of privacy protection are provided. We introduce the privacy operating characteristic (POC) to quantitatively show the resulting trade-off between privacy and security. From a POC, policy makers can select the appropriate operating point for surveillance systems with regard to privacy.

In the data-driven healthcare sector, balancing privacy protection and model performance is critical. This paper enhances accuracy and reliability in survival analysis by integrating differential privacy, deep learning, and the Cox proportional hazards model within a federated learning framework. Traditionally, differential privacy’s noise injection often degrades model performance. To address this, we propose two adaptive privacy budget allocation strategies considering weight changes across neural network layers. The first, LS-ADP, utilizes layer sensitivity to assess the influence of individual layer weights on model performance and develops an adaptive differential privacy algorithm. The second, ROW-DP, comprehensively assesses weight variations and absolute values to propose a random one-layer weighted differential privacy algorithm. These algorithms provide differentiated privacy protection for various weights, mitigating privacy leakage while ensuring model performance. Experimental results on simulated and clinical datasets demonstrate improved predictive performance and robust privacy protection.