Effective Privacy Protection Research Articles

Supervised and Revocable Decentralized Identity Privacy Protection Scheme

Decentralized identity represents an innovative approach based on blockchain to achieve effective identity management. This method utilizes decentralized identifiers and verifiable credential to enable trusted authentication, free circulation of identity information, and self-sovereign control over identity data functionalities. The current decentralized identity systems rely on entirely anonymous identifiers, lacking robust identity regulation. Furthermore, they face challenges such as identity attribute leakage during verifiable credential presentation and the issuers' struggle to reliably revoke credentials. To address these issues, efficient and practical schemes have been designed based on BBS signature, zero-knowledge proof, dynamic accumulator and blockchain technology: one for decentralized identifiers management and the other for verifiable credential privacy protection, both of which are supervised and revocable. The former ensures the privacy of subject identity while achieving regulatability and revocability of identity data by regulator. The latter facilitates selective disclosure of anonymous credentials and reliable revocation. A security analysis shows that the proposed scheme meets anonymity, non-forgeability, regulatory reliability, and revocability reliability, and offers comprehensive and effective privacy protection measures. The experimental results demonstrate that the algorithms designed operate at a millisecond level, which satisfies the demands of blockchain identity management scenarios.

Trajectory privacy preservation model based on LSTM-DCGAN

Rapid scientific and technological development has brought many innovations to electronic devices, which has greatly improved our daily lives. Nowadays, many apps require the permission to access user location information, causing the concern on user privacy and making it an important task to protect user trajectory information. This paper proposes a novel model called LSTM-DCGAN by integrating LSTM (Long Short-Term Memory Network) with DCGAN (Deep Convolution Generative Adversarial Network). LSTM-DCGAN takes the advantages of LSTM to remember attributes in the trajectory data and the generator and the discriminator in DCGAN to generate and discriminate the trajectories. The proposed model is trained using real user trajectory data and the experimental results are validated from the perspectives of both effectiveness and practicality. Results show that the proposed LSTM-DCGAN model outperforms similar methods in generating synthesized trajectories that are similar to real trajectories in terms of the temporal and the spatial characteristics. In addition, various influencing factors are evaluated to investigate ways of further improving and optimizing the model. Overall, the proposed LSTM-DCGAN model can achieve the balance between the effectiveness of privacy protection and the practicality of user trajectory data and can thus be applied to safeguarding user trajectory information.

Privacy protection of communication networks using fully homomorphic encryption based on network slicing and attributes

At present, social networks have become an indispensable medium in people's daily life and work. However, concerns about personal privacy leakage and identity information theft have also emerged. Therefore, a communication network system based on network slicing is constructed to strengthen the protection of communication network privacy. The chameleon hash algorithm is used to optimize attribute-based encryption and enhance the privacy protection of communication networks. On the basis of optimizing the combination of attribute encryption and homomorphic encryption,, a communication network privacy protection method using homomorphic encryption for network slicing and attribute is designed. The results show that the designed network energy consumption is low, the average energy consumption calculation is reduced by 8.69%, and the average energy consumption calculation is reduced by 14.3%. During data transmission, the throughput of the designed network can reach about 700 Mbps at each stage, which has a high efficiency.. The above results demonstrate that the designed communication network provides effective privacy protection. Encrypted data can be decrypted and tracked in the event of any security incident. This is to protect user privacy and provide strong technical support for communication network security.

Effective Privacy Protection Strategies for Pregnancy and Gestation Information From Electronic Medical Records: Retrospective Study in a National Health Care Data Network in China.

Pregnancy and gestation information is routinely recorded in electronic medical record (EMR) systems across China in various data sets. The combination of data on the number of pregnancies and gestations can imply occurrences of abortions and other pregnancy-related issues, which is important for clinical decision-making and personal privacy protection. However, the distribution of this information inside EMR is variable due to inconsistent IT structures across different EMR systems. A large-scale quantitative evaluation of the potential exposure of this sensitive information has not been previously performed, ensuring the protection of personal information is a priority, as emphasized in Chinese laws and regulations. This study aims to perform the first nationwide quantitative analysis of the identification sites and exposure frequency of sensitive pregnancy and gestation information. The goal is to propose strategies for effective information extraction and privacy protection related to women's health. This study was conducted in a national health care data network. Rule-based protocols for extracting pregnancy and gestation information were developed by a committee of experts. A total of 6 different sub-data sets of EMRs were used as schemas for data analysis and strategy proposal. The identification sites and frequencies of identification in different sub-data sets were calculated. Manual quality inspections of the extraction process were performed by 2 independent groups of reviewers on 1000 randomly selected records. Based on these statistics, strategies for effective information extraction and privacy protection were proposed. The data network covered hospitalized patients from 19 hospitals in 10 provinces of China, encompassing 15,245,055 patients over an 11-year period (January 1, 2010-December 12, 2020). Among women aged 14-50 years, 70% were randomly selected from each hospital, resulting in a total of 1,110,053 patients. Of these, 688,268 female patients with sensitive reproductive information were identified. The frequencies of identification were variable, with the marriage history in admission medical records being the most frequent at 63.24%. Notably, more than 50% of female patients were identified with pregnancy and gestation history in nursing records, which is not generally considered a sub-data set rich in reproductive information. During the manual curation and review process, 1000 cases were randomly selected, and the precision and recall rates of the information extraction method both exceeded 99.5%. The privacy-protection strategies were designed with clear technical directions. Significant amounts of critical information related to women's health are recorded in Chinese routine EMR systems and are distributed in various parts of the records with different frequencies. This requires a comprehensive protocol for extracting and protecting the information, which has been demonstrated to be technically feasible. Implementing a data-based strategy will enhance the protection of women's privacy and improve the accessibility of health care services.

Legal Challenges and Countermeasures for Data Privacy Protection

In the digital age, data privacy protection is facing legal challenges and countermeasures. Traditional privacy protection laws have limitations and cannot fully adapt to the era of big data. The illegal collection and utilization of personal information, the failure to meet ethical standards for privacy protection under new technologies, and the lagging legislative practice of personal information protection are all issues that need to be addressed. This article adopts a series of measures and advocates public awareness to enhance citizens' awareness and participation in personal information protection. From an international perspective, the experience of the EU's "data protection" system is worth learning from, while personal information protection involves basic human rights and should be ensured by legislation. In the future, it is necessary to design a path that balances information freedom and privacy protection, explore technology and incentive paths, industry self-discipline and rights building paths, in order to ensure effective protection of data privacy.

Methods and Challenges of Cryptography-Based Privacy-Protection Algorithms for Vehicular Networks

With the rapid development of wireless communication technology, positioning technology, and modern smart devices, Internet of Vehicles (IoVs) smart vehicles have brought great convenience to human production and life. Meanwhile, privacy and security issues are becoming extremely serious, with serious consequences if sensitive data such as vehicle location and trip patterns are leaked. This paper focuses on the demands for vehicular network security, especially privacy protection and existing privacy-protection techniques, including common cryptography methods and cryptography-based advanced technologies. At the same time, this paper also analyzes the advantages and challenges of these technologies in protecting privacy and network security in the Internet of Vehicles, such as the challenges of computational resource requirements and security efficiency in the implementation process, as well as the complexity of realizing effective privacy protection in the interactions among different entities. Finally, this paper envisions the development of privacy-preserving application scenarios and the prospects for crypotography-based privacy-preserving technologies.

Trajectory privacy protection method based on sensitive semantic location replacement

As Location-Based Services (LBSs) proliferate within the Social Internet of Vehicles (SIoVs), the collection and utilization of vehicle trajectories, which are rich in spatio-temporal and semantic informations, have intensified. Publishing these trajectories without adequate privacy safeguards significantly threatens user semantic trajectory privacy. Current methodologies for protecting semantic trajectory privacy do not adequately consider the similarity in query probability between sensitive and alternative semantic locations, thus undermining the efficacy of privacy protection. To address this critical gap, we introduces a novel trajectory privacy protection method based on sensitive semantic location replacement (SSLR), which strategically replaces sensitive semantic locations. This method consists of three key steps: first, semantically annotating trajectory sampling locations to identify sensitive semantic points; second, employing a unique double semicircular area to define replacement regions and selecting replacement Points of Interests (PoIs) that align in semantic attributes and query probabilities; third, reconstructing and publishing the modified trajectories. Our simulation results confirm that SSLR advances the state of the art by delivering superior performance in average recognition rate, geographic similarity, and semantic similarity compared to existing methods. Crucially, by incorporating considerations of query probability similarities, this paper not only enhances the effect of trajectory privacy protection but also preserves its utility. This dual enhancement represents a groundbreaking approach to trajectory privacy, with significant implications for advancing privacy strategies in LBSs within SIoVs.

Exploring the tradeoff between data privacy and utility with a clinical data analysis use case

BackgroundSecuring adequate data privacy is critical for the productive utilization of data. De-identification, involving masking or replacing specific values in a dataset, could damage the dataset’s utility. However, finding a reasonable balance between data privacy and utility is not straightforward. Nonetheless, few studies investigated how data de-identification efforts affect data analysis results. This study aimed to demonstrate the effect of different de-identification methods on a dataset’s utility with a clinical analytic use case and assess the feasibility of finding a workable tradeoff between data privacy and utility.MethodsPredictive modeling of emergency department length of stay was used as a data analysis use case. A logistic regression model was developed with 1155 patient cases extracted from a clinical data warehouse of an academic medical center located in Seoul, South Korea. Nineteen de-identified datasets were generated based on various de-identification configurations using ARX, an open-source software for anonymizing sensitive personal data. The variable distributions and prediction results were compared between the de-identified datasets and the original dataset. We examined the association between data privacy and utility to determine whether it is feasible to identify a viable tradeoff between the two.ResultsAll 19 de-identification scenarios significantly decreased re-identification risk. Nevertheless, the de-identification processes resulted in record suppression and complete masking of variables used as predictors, thereby compromising dataset utility. A significant correlation was observed only between the re-identification reduction rates and the ARX utility scores.ConclusionsAs the importance of health data analysis increases, so does the need for effective privacy protection methods. While existing guidelines provide a basis for de-identifying datasets, achieving a balance between high privacy and utility is a complex task that requires understanding the data’s intended use and involving input from data users. This approach could help find a suitable compromise between data privacy and utility.

Poincaré Differential Privacy for Hierarchy-Aware Graph Embedding

Hierarchy is an important and commonly observed topological property in real-world graphs that indicate the relationships between supervisors and subordinates or the organizational behavior of human groups. As hierarchy is introduced as a new inductive bias into the Graph Neural Networks (GNNs) in various tasks, it implies latent topological relations for attackers to improve their inference attack performance, leading to serious privacy leakage issues. In addition, existing privacy-preserving frameworks suffer from reduced protection ability in hierarchical propagation due to the deficiency of adaptive upper-bound estimation of the hierarchical perturbation boundary. It is of great urgency to effectively leverage the hierarchical property of data while satisfying privacy guarantees. To solve the problem, we propose the Poincar\'e Differential Privacy framework, named PoinDP, to protect the hierarchy-aware graph embedding based on hyperbolic geometry. Specifically, PoinDP first learns the hierarchy weights for each entity based on the Poincar\'e model in hyperbolic space. Then, the Personalized Hierarchy-aware Sensitivity is designed to measure the sensitivity of the hierarchical structure and adaptively allocate the privacy protection strength. Besides, Hyperbolic Gaussian Mechanism (HGM) is proposed to extend the Gaussian mechanism in Euclidean space to hyperbolic space to realize random perturbations that satisfy differential privacy under the hyperbolic space metric. Extensive experiment results on five real-world datasets demonstrate the proposed PoinDP’s advantages of effective privacy protection while maintaining good performance on the node classification task.

PDMSC: privacy-preserving decentralized multi-skill spatial crowdsourcing

Purpose The purpose of this paper is to study the following two issues regarding blockchain crowdsourcing. First, to design smart contracts with lower consumption to meet the needs of blockchain crowdsourcing services and also need to design better interaction modes to further reduce the cost of blockchain crowdsourcing services. Second, to design an effective privacy protection mechanism to protect user privacy while still providing high-quality crowdsourcing services for location-sensitive multiskilled mobile space crowdsourcing scenarios and blockchain exposure issues. Design/methodology/approach This paper proposes a blockchain-based privacy-preserving crowdsourcing model for multiskill mobile spaces. The model in this paper uses the zero-knowledge proof method to make the requester believe that the user is within a certain location without the user providing specific location information, thereby protecting the user’s location information and other privacy. In addition, through off-chain calculation and on-chain verification methods, gas consumption is also optimized. Findings This study deployed the model on Ethereum for testing. This study found that the privacy protection is feasible and the gas optimization is obvious. Originality/value This study designed a mobile space crowdsourcing based on a zero-knowledge proof privacy protection mechanism and optimized gas consumption.

Model-level attention and batch-instance style normalization for federated learning on medical image segmentation

Federated learning (FL) offers an effective privacy protection mechanism for cross-center medical collaboration and data sharing. In multi-site medical image segmentation, FL allows each medical site to act as a client, forming its own data domain. FL has the potential to enhance the performance of models on known domains. However, practical deployment faces the challenge of domain generalization (DG) due to the non-identical and non-independent (non-IID) nature of data from different domains. This results in decreased model performance in unseen domains. Current DG solutions are overly complex in addressing style differences and lack focus on inter-domain image features causing model differences. Furthermore, these solutions are not suitable for the FL paradigm that requires data storage separation. Hence, the lightweight model-level attention and batch-instance style normalization (MLA–BIN) is proposed to solve the DG of FL in this study. The MLA module represents the unseen domain as a linear combination of seen domain models. It does not require access to raw data but learns from the sufficient exploration of data features in known domains, thereby identifying differences in inter-domain data features and enabling the global model to generalize from seen to unseen domains. In the BIN block, batch normalization (BN) and instance normalization (IN) are combined to perform the shallow layers of the segmentation network for style normalization. By integrating the segmentation backbone network with the BIN block (BIN–Net), it ensures effective learning of intra-domain features and addresses the impact of inter-domain image style differences on domain generalization without accessing data from other centers. Extensive experimental results demonstrate that the proposed method achieved a Dice similarity coefficient of 88.27, 88.25 and 64.94 on the prostate, the optic disc and cup, and the COVID-19 lesion segmentation dataset, respectively, outperforming the state-of-the-art methods.

Multidimensional Epidemiological Survey Data Aggregation Scheme Based on Personalized Local Differential Privacy

In recent years, with the rapid development of intelligent technology, information security and privacy issues have become increasingly prominent. Epidemiological survey data (ESD) research plays a vital role in understanding the laws and trends of disease transmission. However, epidemiological investigations (EI) involve a large amount of privacy-sensitive data which, once leaked, will cause serious harm to individuals and society. Collecting EI data is also a huge task. To solve these problems and meet personalized privacy protection requirements in EIs, we improve the uOUE protocol based on utility-optimized local differential privacy to improve the efficiency and accuracy of data coding. At the same time, aiming at the collection and processing of ESD, a multidimensional epidemiological survey data aggregation scheme based on uOUE is designed. By using Paillier homomorphic encryption and an identity-based signature scheme to further prevent differential attacks and achieve multidimensional data aggregation, the safe, efficient, and accurate aggregation processing of ESD is executed. Through security proof and performance comparison, it is verified that our algorithm meets the requirements of local differential privacy and unbiased estimation. The experimental evaluation results on two data sets show that the algorithm has good practicability and accuracy in ESD collection and provides reliable and effective privacy protection.

ES Data Aggregation Scheme based on Personalized Local Differential Privacy

In modern society, with the rapid development of technology, research on Epidemiological Survey Data (ESD) has become increasingly crucial. As various diseases continue to evolve and spread, there is a growing need for a profound understanding of the patterns and trends of disease transmission. Epidemiological investigation, as a key method, provides fundamental data support for devising effective prevention and control strategies by tracking cases, contacts, and potential infectees. However, accompanying this progress is the issue of dealing with a large amount of privacy-sensitive data. Once these data are compromised, it may pose severe harm to individuals and society. To address this challenge, we propose a uOUE-based epidemiological survey data aggregation scheme for the collection and processing of ESD, aiming to enhance the efficiency and accuracy of data coding. We ensure the secure, efficient, and accurate aggregation processing of ESD. Through rigorous validation and comparative analysis, our algorithm complies with the requirements of local differential privacy and unbiased estimation. It demonstrates good practicality and accuracy in ESD collection, providing users with a reliable privacy protection effect.

Population Neuroscience: Strategies to Promote Data Sharing While Protecting Privacy.

Population neuroscience aims to advance our understanding of how genetic and environmental factors influence brain development and brain health over the life span, by integrating genomics, epidemiology, and neuroscience at population scale. This big data approach depends on data sharing strategies at both the micro- and macro-level, as well as attention to effective data management and protection of participant privacy. At the micro-level, researchers participate in international consortia that support collaboration, standards, and data sharing. They also seek to link together cohort studies, administrative health databases, and measures of the physical, built, and social environment in creative ways. Large-scale, longitudinal, and multi-modal cohorts are being designed to support explorations of genetic and environmental impacts on the brain. At a macro-level, funding agency policies now require data across health research domains to be managed according to the FAIR (findable, accessible, interoperable, and re-useable) Data principles and made available to the research community in a timely manner to support reproducibility and re-use. Data repositories provide technical infrastructure for storing, accessing, and increasingly also analyzing rich population-level data. Federated and cloud-based approaches are being leveraged to improve the security, remote accessibility, and performance of repositories. Finally, legal frameworks are being developed to facilitate secure health data access, integration, and analysis, providing new opportunities for the field.

Challenges for children’s rights in connection with the development of artificial intelligence

This article explores the importance of protecting children’s rights in the context of the rapid development of artificial intelligence. Artificial intelligence has the potential to change many aspects of a child’s life, and while it can have positive effects, its negative impact also needs due consideration.
 The article focuses on four key aspects: privacy, security, discrimination and ethics. It analyzes the risks associated with the collection and processing of children’s personal data by artificial intelligence and requires the establishment of effective privacy protection mechanisms. In addition, it considers the safety of children, especially in the context of the use of autonomous robots and toys, and safety standards that are required to be taken into account in the development of such systems.
 The article also focuses on the problem of discrimination that may arise from the systematic use of artificial intelligence. It demands developers to ensure the fairness of algorithms and avoid discrimination against children. Finally, the article considers the ethical aspects of the use of artificial intelligence, namely the question of the responsibility of developers and the need for ethical principles in all aspects of its use.
 In conclusion, the article emphasizes the need for constant monitoring and regulation of the development of artificial intelligence in order to protect the rights of the child. It recommends the implementation of an effective legislative framework that defines standards for the protection of children’s rights in the context of artificial intelligence. In addition, the article puts forward the idea of developing ethical directives and codes of conduct for developers, researchers and users of artificial intelligence from the perspective of children’s rights.
 In general, this scientific article emphasizes the importance of protecting the rights of the child in the context of the development of artificial intelligence. It calls for action at the level of legislation, technological standards and ethical principles, ensuring safety, privacy, non-discrimination and ethical responsibility in all aspects of the use of artificial intelligence to ensure the healthy development and protection of children’s rights in the digital age.
 Also, the article is devoted to the analysis of the controversial issue of the relationship between child rights and artificial intelligence. The article examines the consequences of the use of artificial intelligence in the context of children’s rights, the main issues related to the legal protection of children in the field of artificial intelligence, and suggests possible solutions for these problems.

A Privacy-Preserving Trajectory Publishing Method Based on Multi-Dimensional Sub-Trajectory Similarities.

With the popularity of location services and the widespread use of trajectory data, trajectory privacy protection has become a popular research area. k-anonymity technology is a common method for achieving privacy-preserved trajectory publishing. When constructing virtual trajectories, most existing trajectory k-anonymity methods just consider point similarity, which results in a large dummy trajectory space. Suppose there are n similar point sets, each consisting of m points. The size of the space is then mn. Furthermore, to choose suitable k- 1 dummy trajectories for a given real trajectory, these methods need to evaluate the similarity between each trajectory in the space and the real trajectory, leading to a large performance overhead. To address these challenges, this paper proposes a k-anonymity trajectory privacy protection method based on the similarity of sub-trajectories. This method not only considers the multidimensional similarity of points, but also synthetically considers the area between the historic sub-trajectories and the real sub-trajectories to more fully describe the similarity between sub-trajectories. By quantifying the area enclosed by sub-trajectories, we can more accurately capture the spatial relationship between trajectories. Finally, our approach generates k-1 dummy trajectories that are indistinguishable from real trajectories, effectively achieving k-anonymity for a given trajectory. Furthermore, our proposed method utilizes real historic sub-trajectories to generate dummy trajectories, making them more authentic and providing better privacy protection for real trajectories. In comparison to other frequently employed trajectory privacy protection methods, our method has a better privacy protection effect, higher data quality, and better performance.

Thematic exploration of online privacy literacy and examination of its future agenda

ABSTRACT In today’s digitized society of ever-increasing information sharing, Online Privacy Literacy (OPL) is not only important to feel safe but be safe as well. OPL refers to skills and knowledge of individuals that helps to protect their privacy online. OPL is crucial in the digital era, where extensive data collection poses privacy challenges. The lack of OPL is a major disadvantage as it may impair one’s privacy decision-making abilities such as whether and how much personal data to disclose online. Despite its importance, OPL research is in its early stages. To pave the way to bridge this gap, we conducted a systematic review of 57 articles, offering a theoretical connection to privacy research. Our synthesis highlights how OPL is defined, measured, and its primary knowledge domains. Seven themes emerged, emphasizing OPL across age groups and revealing unexplored areas like the interplay between OPL, technology acceptance, and technostress. This exploration provides a roadmap for researchers to understand OPL implications and its role in privacy domains, guiding future research directions. Experts can leverage this thematic exploration to develop knowledge, skills, and effective privacy protection measures for individuals and institutions in the dynamic landscape of online privacy.

Hashing Fake: Producing Adversarial Perturbation for Online Privacy Protection Against Automatic Retrieval Models

The wide application of deep neural networks (DNNs) has significantly improved the performance of hashing models on multimodal retrieval issues. DNN-based deep models can automatically learn semantic features from raw data to make human-level decisions. However, the superior generalization leads to potential privacy leakage risks. Strong DNN-based retrieval models enable malicious crawlers to search for nontag private information based on semantic similarity matching. Hence, executing effective privacy protection mechanisms against those retrieval software is essential for reliable social website construction. In this article, we propose a retrieval task-based adversarial perturbation generation method called Hashing Fake to meet this request. Specifically, DNNs are recently found to be vulnerable to a specific set of attacks called adversarial perturbations, which denote some magnitude-restricted signals added on objective samples to misguide well-crafted DNN models, and perturbations’ magnitudes are small enough that will not induce humans’ perception. Moreover, since existing adversarial perturbation generation methods are designed for supervised tasks, Hashing Fake constructs a differential approximation substitution for perturbation production on unsupervised retrieval tasks. Through extensive experiments on several deep retrieval benchmarks, we demonstrate that well-crafted perturbations using Hashing Fake can effectively misguide objective models’ recognitions to make false predictions. The added norm-restricted perturbations on objective samples will not alter humans’ perception; hence, Hashing Fake can be applied on real-world social websites to protect subscribers’ privacy against malicious retrieval software.

Location Privacy Protection and Coverage Hole Effects in Event Monitoring Wireless Networks for Internet of Things Applications

Source-location privacy (SLP) protection improves security in event monitoring wireless sensor networks (WSNs) for Internet of Things (IoT) applications. However, due to constrained energy resources, WSNs incur coverage holes that affect the level of SLP protection. Existing studies have ineffectively analysed the effects of coverage holes on SLP protection. Noting the limitation, this study investigated the coverage hole effects. Performance of various SLP routing protocols was evaluated based on the sensor node utilization ratio, attack success rate, end-to-end delay, and packet delivery ratio. Then, considering the challenge of limited battery power in IoT sensors, the performance gains and limitations of the protocols were identified. Simulation results demonstrate that the data dissemination SLP routing protocol (DIRP) outperforms other protocols. However, the performance of DIRP is significantly affected by coverage holes. The results suggest that for DIRP to be viable in IoT applications, the integration of distributed energy resources should be considered.

A Communication-Efficient, Privacy-Preserving Federated Learning Algorithm Based on Two-Stage Gradient Pruning and Differentiated Differential Privacy

There are several unsolved problems in federated learning, such as the security concerns and communication costs associated with it. Differential privacy (DP) offers effective privacy protection by introducing noise to parameters based on rigorous privacy definitions. However, excessive noise addition can potentially compromise the accuracy of the model. Another challenge in federated learning is the issue of high communication costs. Training large-scale federated models can be slow and expensive in terms of communication resources. To address this, various model pruning algorithms have been proposed. To address these challenges, this paper introduces a communication-efficient, privacy-preserving FL algorithm based on two-stage gradient pruning and differentiated differential privacy, named IsmDP-FL. The algorithm leverages a two-stage approach, incorporating gradient pruning and differentiated differential privacy. In the first stage, the trained model is subject to gradient pruning, followed by the addition of differential privacy to the important parameters selected after pruning. Non-important parameters are pruned by a certain ratio, and differentiated differential privacy is applied to the remaining parameters in each network layer. In the second stage, gradient pruning is performed during the upload to the server for aggregation, and the final result is returned to the client to complete the federated learning process. Extensive experiments demonstrate that the proposed method ensures a high communication efficiency, maintains the model privacy, and reduces the unnecessary use of the privacy budget.