216 results about "Differential privacy" patented technology

The application discloses a kind of based on multi-scale control technology's unmanned system cooperation privacy protection method, belong to unmanned system safety cooperation control field.This method solves the ill-conditioned numerical problem in the cooperative control of unmanned system with fast-slow coupling characteristics, and the state privacy leakage problem caused by communication eavesdropping.Scheme includes: establishing system dynamic model;Design the differential privacy leader-following multi-scale cooperative controller containing attenuation noise;Obtain the standard singular perturbation form of error system through matrix transformation;Based on multi-scale control theory, solve the state control gain that guarantees mean square consistency;Determine the noise control gain and privacy parameter in combination with probability theory and matrix spectral theory.The present application avoids the ill-conditioned numerical problem, and the controller is robust, while achieving accurate cooperative control, provides verifiable differential privacy protection for the initial state of the system, significantly improves the safety and reliability of unmanned system in the hostile environment.

The application discloses a kind of personalized medication service data security processing systems of fusing zero trust architecture and differential privacy, it is related to medical data security technical field, including dynamic trust evaluation to request subject;Inject differential privacy noise;Combining zero trust strategy continues to obtain trust score;Only receive desensitized statistical information or encrypted intermediate representation in the cloud.The application solves the privacy leakage and trust loss problems caused by static protection and weak desensitization in traditional personalized medication services, significantly improves the security, compliance and user trustworthiness of medical data during collection, use and sharing, and provides a secure foundation that can be implemented, verified and continuously evolved for intelligent medication services in high-sensitivity scenarios.

The application relates to a power grid production data access method and device for power grid data sharing security compliance based on deep action determination and risk budget. The method comprises the following steps: in response to an access request for power grid production data sent by a terminal, inputting a context feature vector corresponding to the access request into a pre-constructed deep action determination model to obtain a target access action corresponding to the access request; in the case that the target access action is a preset access action, determining a current privacy budget value corresponding to the target access action; in the case that the current privacy budget value does not exceed a current remaining privacy budget value, performing differential privacy protection processing on the power grid production data according to the current privacy budget value to obtain processed power grid production data, and returning the processed power grid production data to the terminal. By using the method, the security of power grid production data access can be improved.

A multi-level privacy protection method based on bimodal PSNR quality control. A bimodal PSNR evaluation system is constructed to separate the calculation of Face-PSNR and Background-PSNR. An adaptive blur kernel is used for dynamic desensitization, and the kernel is calculated as max(w, h) / 6 based on the ROI size. A PSNR-driven iterative blur mechanism is introduced, and each round is checked until it meets the standard and is backed up with a mosaic. Differential privacy protection is applied to the log, and a heterogeneous detector is used to verify the desensitization effect independently. An accuracy attenuation check and compliance audit mechanism is established. This method achieves real-time processing at 17.8 FPS on Raspberry Pi 4B, with a desensitization incremental delay of 3.6 ms, ROIPSNR<22 dB and global PSNR>35 dB, accuracy attenuation≤3.41%, and privacy unidentifiability rate of 100%, meeting the GDPR / PIPL compliance requirements. It is suitable for privacy protection target detection in community security edge computing scenarios.

This invention relates to a decentralized federated learning optimization method and system based on a consensus Kalman filter, belonging to the fields of privacy computing and data security. The invention aims to optimize and improve the effectiveness of decentralized federated learning methods based on differential privacy. It comprises two core components: a gradient filtering algorithm based on a consensus Kalman filter and an overshoot filtering algorithm based on a consensus Kalman filter. The former uses a Kalman filter to filter gradient information during model training, reducing noise bias in the gradient, increasing gradient accuracy and stability, and thus accelerating the model's convergence speed and accuracy. The latter addresses the overshoot problem caused by momentum optimization methods by using a Kalman filter with exponentially decaying system variance to filter overshoot noise in the model parameters, mitigating the impact of overshoot on model convergence speed and accuracy, ensuring accelerated model convergence in the early stages and stable model convergence in the later stages, thereby achieving higher accuracy.

This invention discloses an intelligent advertising design and generation method, specifically including the following steps: S1, Data Collection and Preprocessing: Collecting product data, user behavior data, historical advertising data, and real-time market data, and performing cleaning, labeling, and feature extraction; S2, User Profile Modeling: Constructing a dynamically updated user interest graph based on clustering analysis and deep learning algorithms; This invention relates to the field of artificial intelligence technology. This intelligent advertising design and generation method integrates multi-dimensional data such as product functions, user behavior, and competitor dynamics to improve the accuracy of advertising content, generate targeted advertising copy, support the joint generation and style unification of text, images, and videos, and the system can simultaneously generate advertising slogans such as "Riding the Wind, Unlocking the Acceleration of Youth" and dynamic posters, track click-through rates, conversion rates, and other indicators in real time, automatically adjust content parameters, use blockchain technology to store copyright information, and protect user data through differential privacy algorithms.

The application discloses a data privacy protection method and a detection method for a privacy-enhanced semi-blind digital fingerprint, combines a differential privacy technology and a digital fingerprint technology, embeds carefully designed noise into specific positions of a numerical aggregate data set, can realize fingerprint embedding and noise interference operation on carrier data in one step, and ensures protection of private information in the carrier data set and tracking function on traitors.

The application belongs to the technical field of information security, and discloses an attribute correlation-based differential privacy anonymous method. First, the correlation between attributes is analyzed through data distribution analysis of a data set, the number of each equivalent group of each attribute group is calculated by using permutation and combination, all minimum attribute combinations are found, if the number of each equivalent group of the attribute combination is greater than or equal to 30, and there is an equivalent group with a number less than 30 in each equivalent group when any attribute is added, the attribute combination is a minimum equivalent group. Then, the set condition of each attribute is obtained by performing intersection on all minimum attribute groups, the number of each attribute under the intersection is calculated, then the priority of the attributes with intersection is divided, the intersection number is used as the privacy protection priority of the attribute, privacy budget is allocated through the privacy protection priority of each attribute, and finally, the attributes with intersection are protected by differential privacy.

The present application relates to the technical field of electronic digital data processing, in particular to a private protection method for sensitive data of a carrier in a smart logistics platform, comprising designing a local differential privacy protection algorithm for multi-dimensional numerical data of the carrier and a local differential privacy protection algorithm for location data of the carrier; the multi-dimensional numerical data of the carrier is disturbed by using the local differential privacy protection algorithm for multi-dimensional numerical data of the carrier, and private protection numerical data is obtained; the location data of the carrier is disturbed by using the local differential privacy protection algorithm for location data of the carrier, and private protection location data is obtained; the multi-dimensional numerical data and the location data of the carrier are respectively protected by the two protection algorithms, and the problem that the existing smart logistics platform does not protect the logistics process data of the carrier and causes the personal privacy of the carrier to be leaked is solved.

A personal data protection method of the present disclosure includes sending, by a user process configured to communicate with a secure process via a remote procedure call (RPC), a first RPC request related to a first operation and a second RPC request related to a second operation to the secure process. The first operation is an operation of reading personal data or an operation of transforming the personal data. The second operation is an operation of applying differential privacy to statistical information based on the personal data. The personal data protection method further includes, by the secure process, executing the first operation based on the first RPC request, returning, to the user process, a handler corresponding to data obtained through the first operation, executing the second operation based on the second RPC request, and returning, to the user process, the statistical information to which the differential privacy has been applied.

Various examples are directed to systems and methods for using a machine learning model. A computing system may access training data comprising a plurality of training data items. Each of the plurality of training data items may comprise a plurality of features. From a first training data item of the plurality of training data items, the computing system may generate a first transformed training data item using a first privacy budget corresponding to a first portion of the first training data item and a second privacy budget corresponding to a second portion of the first training data item. The computing system may train a machine learning model using the first transformed training data item and use the trained machine learning model to generate at least one class probability for a data item.

A method and system for low-rank adaptation and privacy protection of a large geospatial model are disclosed, involving large geospatial models and federated learning techniques. The method includes: loading a pre-trained large geospatial model and inserting a low-rank adapter, freezing the backbone and training only the adapter; extracting features based on local remote sensing ecological monitoring data and calculating local feature prototypes; performing hierarchical differential privacy and communication compression on the adapter update and prototype before uploading; a central server aggregates and generates an initial global prototype, adaptively assigning weights based on the differences between the local and global prototypes, and weighted aggregation to obtain the global model and the updated global feature prototype; distributing the global model and performing prototype alignment regularization training with the global prototype as a constraint to achieve alignment between the local and global feature spaces. The system includes modules for local training, privacy communication, global aggregation, and distribution alignment. This invention protects data privacy, reduces communication overhead, and improves cross-node generalization ability, making it suitable for distributed geospatial ecological monitoring in weak network environments.

The application provides a differential privacy protection method and device for federated learning, which comprises the following steps: obtaining model weight differences uploaded by each client participating in current round learning; performing clipping operation on the model weight differences uploaded by each client according to the clipping parameter corresponding to the current round learning; aggregating each model weight difference after the clipping operation, and performing noise adding processing on the aggregated model weight difference according to the Gaussian noise distribution corresponding to the current round learning to complete model updating of the current round learning; wherein the Gaussian noise distribution corresponding to the current round learning is determined according to the noise scale corresponding to the current round learning and the clipping parameter corresponding to the current round learning, and the noise scale corresponding to each round of learning gradually decreases with the increase of the learning round. The added noise can be fitted to the characteristics of the model weight information uploaded by the current client, thereby obtaining higher model precision and effectively reducing the privacy budget in differential privacy protection.

The application provides a big data federal learning analysis method and system considering privacy protection, and belongs to the field of big data processing and data security. In view of the problems of traditional centralized analysis, such as easy data leakage, low compliance, and the problems of existing federal learning, such as insufficient privacy protection, low model accuracy, and untraceable parameter transmission, a scheme combining local hierarchical encryption desensitization, differential privacy disturbance, weighted federal aggregation and blockchain evidence is adopted. Each node encrypts and desensitizes the data according to the sensitivity level, locally completes model training, encrypts and transmits the disturbed parameters, and uploads and stores the evidence on the chain. The coordination node allocates weights according to the data volume and quality of each node, weighted aggregation generates a global model and iteratively optimizes it. The method realizes data "available but invisible", while ensuring privacy security and compliance, improves analysis accuracy and training efficiency, and can be used for multi-source sensitive data cross-domain collaborative analysis in medical treatment, finance, government affairs and the like.

The application discloses a kind of multi-source consumption behavior personalized recommendation method and system for privacy computing, it is related to data processing technical field, including, using dynamic differential privacy disturbance algorithm, to structured feature vector set is carried out differential privacy disturbance processing, generates differential privacy feature data, and carries out homomorphic encryption, generates encrypted feature dataset;Iterative training is carried out to encrypted feature dataset, generates local encrypted parameter;Local encrypted parameter is transmitted to encryption aggregation center, obtains encrypted parameter set, and using homomorphic encryption aggregation method, encrypted parameter set is merged, generates global encrypted recommendation parameter;Global encrypted recommendation parameter is decrypted in trusted execution environment, obtains available recommendation parameter, and combines local encrypted parameter and differential privacy feature data and carries out recommendation calculation, generates personalized recommendation result.The application improves privacy protection capability and data availability.

The application provides a power edge lightweight privacy protection method and system supporting a national encryption algorithm, wherein the national encryption algorithm supported by a power edge node is used for initializing a random number generator, ensuring that the randomness of a full homomorphic encryption algorithm calculation bottom layer meets the national encryption standard, realizing the compatibility of the two, guaranteeing the safe and trusted transmission of federal learning data, introducing a mean square error gradient can avoid the ciphertext domain nonlinear calculation of the full homomorphic encryption algorithm, adapting to the situation that the power edge node resources are limited, and introducing a dynamic differential privacy noise mechanism considering annealing attenuation and node cooperative regulation can avoid the inverse decryption analysis of the initiator on the global characteristics of the data, realizing the blocking of data reconstruction attacks while guaranteeing the precision and convergence efficiency of the local model, so as to better support the calculation demand of the power edge scene.

A wireless communication network method and system that includes a central server, cellular base stations, edge computing devices, and client devices. The client devices send uplink pilot sequences, which are collected and aggregated by the base stations. The base stations then relay this aggregated pilot data to the central server. The central server deploys network weights for a global deep learning neural network model to the base stations for incorporation into respective local deep learning models, trained to predict optimal beamforming vectors, to the base stations for incorporation into respective local deep learning model. During training the central server integrates adaptive noise into weights received for each of the local deep learning models.

This invention provides a local differential privacy collaborative sampling data collection method and system resistant to poisoning attacks, belonging to the field of information security and privacy protection technology. The scheme compares the single-dimensional probability distribution reported by the user terminal with the median probability distribution to identify and eliminate abnormal reported data, thereby effectively suppressing data poisoning attacks initiated by malicious user terminals. Subsequently, under the constraint of local differential privacy, combined with a generalized random response mechanism and a collaborative sampling strategy, multidimensional data is sampled differentially to achieve more accurate frequency estimation. The scheme of this invention improves the privacy protection strength and the accuracy of statistical results in multidimensional data collection without relying on a trusted third party.

This invention relates to the field of intelligent education technology, specifically to a method and system for guiding student device self-management based on achievement incentives. The method includes: performing differential privacy perturbation processing on raw data of student device usage behavior to form aggregated statistical behavior data; associating and mapping the aggregated statistical behavior data with a preset educational scenario template to generate an initial control strategy for student devices; deploying a process protection component according to the initial control strategy to generate process protection status data; merging the process protection status data and the aggregated statistical behavior data, and generating a self-management badge and matching corresponding self-management permissions when the behavior performance evaluation meets the continuous achievement conditions; after a student's temporary time extension request initiated by the permission is approved, optimizing the control strategy by coupling peripheral device status data, synchronizing it to the local policy engine, and supporting the independent adjustment of some parameters; this invention can improve the efficiency of student device self-management.

This invention discloses a blockchain-based method for trusted analysis of multi-party data sharing, involving the integrated application of blockchain technology, privacy computing technology, and distributed data collaboration technology. The method constructs a two-way auction market for computing power, employs an optimistic game theory mechanism to schedule tasks, utilizes an adaptive privacy-based routing mechanism based on information entropy, and dynamically matches fully homomorphic encryption, TEE, or differential privacy computing paths according to data sensitivity. It introduces a random arbitration mechanism based on verifiable random functions to handle disputes and combines a Shapley value-driven zero-knowledge contribution measurement mechanism for incentive allocation. This invention breaks the traditional full-participant verification paradigm, solves the problems of wasted privacy computing resources and unfair incentives, and significantly improves the system's throughput, real-time performance, and security.

This invention relates to the field of vehicle network data security technology, specifically to a vehicle network data sharing method based on differential privacy-enhanced federated learning. For multi-source heterogeneous data sharing scenarios in vehicle networks, it constructs a bidirectional dynamic privacy protection mechanism across the entire link of local model upload and global model distribution, enabling secure collaborative training where the model moves while the data remains stationary. Hardware-level algorithm optimization perfectly adapts to vehicle computing power constraints. This invention addresses the computing power limitations of ARM-based vehicle devices by employing adaptive noise injection to achieve an optimal balance between privacy protection and model accuracy. Based on a model parameter sensitivity grading strategy, this invention is highly efficient and low-latency, adaptable to large-scale vehicle network applications. Through optimization strategies such as edge pre-aggregation, sparsity processing, and pipelined parallel computing, this invention significantly reduces computational latency and communication transmission volume.

The application provides a direction perception-based trajectory data collection method and system satisfying local differential privacy, first, through an anchor-based spatial constraint method, the trajectory area of a user is adaptively limited, so that the performance is significantly improved without violating the privacy constraint; second, the adjacent direction information of each point in the trajectory is discretized for the disturbance process of the point by using a privacy budget; the information is used as a clue for connecting adjacent points, and can be used to limit the area of each point in the trajectory; finally, the final disturbed trajectory is obtained by using an exponential mechanism and an optimization process and is uploaded to a server; the application satisfies strict local differential privacy, provides provable privacy protection for mobile user trajectory data, and does not need to access additional public knowledge, and solves the privacy leakage problem in the trajectory data collection process by using the method.

This invention relates to the field of cloud computing technology, and particularly to a cross-domain data collaborative computing method based on secure multi-party computation and differential privacy. The invention proposes the following scheme: the task initiator abstracts the collaborative task into an intent description object and compiles it into an intermediate representation with gating logic, generating a dense-state function load for joint evaluation by multiple parties; the task assisting end generates real local contributions and counterfactual comparison contributions locally based on private data, and forms a detection score according to the request trajectory, adaptively determining the mixing coefficients. The two types of contributions are then subjected to restricted mixing and direction correction processing before participating in the dense-state computation. This application, without changing the appearance of the evaluation protocol, balances task intent protection and sensitive subgroup existence protection, improving security and engineering usability in cross-domain collaborative scenarios.

This invention belongs to the field of data transmission technology and discloses a privacy-preserving network security data transmission method, including the following steps: S1. Data privacy preprocessing: Constructing a personalized differential privacy protection mechanism to perform local privatization processing on the original data to balance data privacy protection and availability; S2. Intelligent scheduling of transmission tasks: For scenarios with multiple network nodes and multiple transmission links, establishing a scheduling model with transmission latency and energy consumption as optimization objectives, and using a quantum particle swarm optimization algorithm that integrates chaotic perturbations to solve the model; S3. Trusted verification of transmitted data: Dividing the network into several sub-regions and constructing a tree-like key management structure, and using key priority allocation and cross-region node verification mechanisms. This solution, through a three-step collaborative mechanism, constructs an end-to-end network security protection system covering the entire data lifecycle, achieving the organic unity of personalized privacy protection, collaborative optimization of transmission efficiency and energy consumption, and decentralized trusted verification.

This invention relates to a recommendation method based on verifiable local differential privacy, belonging to the field of privacy protection technology. The specific process of this method is as follows: the client generates a client-side random factor; the client uses a server-side random seed to generate a server-side random factor; the client-side random factor and the server-side random factor are combined to obtain a perturbation factor; an interaction rating vector is generated based on the client's interaction with the product; the perturbation factor is used to perform random perturbation on the interaction rating vector to obtain a perturbation rating vector that satisfies local differential privacy; the client constructs a zero-knowledge proof, which proves that the client strictly follows a predefined random perturbation mechanism; the server verifies the zero-knowledge proof and aggregates the perturbation rating vectors of all clients that pass the verification; a recommendation list is generated based on the aggregation result, and the recommendation list is returned to all clients.