52 results about "Privacy model" patented technology

Embodiments of the present invention address deficiencies of the art in respect to privacy compliance assessment for computer software and provide a method, system and computer program product for a privacy model framework for software applications. In one embodiment, a privacy modeling data processing system can be provided. The privacy modeling data processing system can include a modeling framework configured for communicative coupling to a software application. The modeling framework can capture information flows from requests to and responses from a coupled software application, and can rules-based process the captured information flows for privacy rules to generate a privacy compliance report for the software application.

A system and method for providing “static analysis” of programs to aid in improving runtime performance, stability, security and privacy characteristics of deployed application code. The method includes performing a set of analyses that sifts through the program code and identifies programming security and / or privacy model coding errors. In particular the invention focuses on identifying coding errors that cause loss of correctness, performance degradation, security, privacy and maintainability vulnerabilities. A deep analysis of the program is performed using detailed control and data flow analyses. These deeper analyses provide a much better perspective of the overall application behavior. This deep analysis is in contrast to shallow analyses in current industry tools, which inspect or model a single or a few classes at a time.

The invention provides a privacy protection method and system based on knowledge migration under a collaborative learning framework. The method comprises the following steps: segmenting a local privacy data set into a plurality of disjoint privacy subsets, and training a corresponding privacy model based on each privacy subset; collaborative learning: submitting the public data to a privacy model,using an aggregation mechanism by the privacy model, and migrating knowledge to a public data set to label a label; after enough labeled data is obtained, a local interaction model is trained, part of parameters are uploaded to a server in each round of training, the server updates and maintains global parameters and provides latest parameters for all parties to download, and participants download the latest parameters to optimize the local interaction model. According to the invention, multiple parameter interactions of collaborative learning can also be ensured, and even if the interactionmodel obtains less annotation data, high accuracy can still be maintained after multiple rounds of training.

The invention discloses a vehicle position privacy protection method of a vehicular ad hoc network. The method comprises the following steps: a privacy model is first set to assess a current position privacy of different vehicles so as to reflect a reputation module and a mix-zone criterion of a cooperation degree of the current vehicle; any one vehicle in the vehicular ad hoc network determines whether to cooperate and change a pseudo name to protect a vehicle i according to a current reputation value of the vehicle i sending a pseudo name request and based on a judgment whether the one vehicle is within a mix-zone of the vehicle i; and the reputation value of the one vehicle can be dynamically adjusted according to the current position privacy of the one vehicle and pseudo name residual time. The method of the invention can meet the requirement of protecting the current position of the vehicle and can substantially reduce the energy cost; and the method has the advantages that more vehicles are involved in the pseudo name change and the safety of the position privacy of the vehicular wireless network is greater.

The invention discloses a vehicle position privacy protection method based on group signature and reputation, which comprises a reputation model for encouraging more vehicles to cooperate with each other to form a group, a privacy model for evaluating the position privacy strength of the vehicle, a group forming and dynamic mixed region establishing mechanism, and a pseudonym replacement mechanism. Any vehicle in the vehicle-mounted ad hoc network actively requests establishment of a group from a control server according to the reputation value of the vehicle when the pseudonym of the vehicleis soon to be expired; the control server allocates a group leader role to the requesting vehicle and dynamically forms a mixed region according to the current position of the requesting vehicle; theother vehicles in the mixed region judge whether the cooperation is participated in or not to change the pseudonym according to the reputation value, privacy strength and the remaining time of the pseudonym of the other vehicles; and the reputation value of the vehicles can be dynamically adjusted. According to the method, more vehicles can be encouraged to join the group, thereby changing the pseudonym in cooperation and ensuring the position privacy of the vehicles; meanwhile, the vehicles do not need to keep a silent state in the formed mixed region but broadcast messages as a group identity so that the running safety of the vehicles is guaranteed; furthermore, resource expenditure caused by unnecessary frequent change of the pseudonym can be effectively reduced.

The invention discloses a graded privacy protection method in hierarchical data release. The graded privacy protection method comprises the following steps: firstly, grading sensitive attributes of original hierarchical data according to actual sensitivity; then, limiting the appearing frequency of the sensitive attributes with different grades in an equivalence class according to a divided gradesetting frequency parameter alphai; then realizing a given privacy model by adopting a clustering method. By adopting the graded privacy protection method disclosed by the invention, the problem of privacy leakage, caused by close-in attack, of the sensitive attributes of an existing hierarchical data release technology is solved very well, the protection intensity on the privacy protection of thehierarchical data is improved and the safety of the data is enhanced; the graded privacy protection method in the hierarchical data release, disclosed by the invention, has certain real meaning and practical value in the protection of the hierarchical data.

The invention provides a spatial aggregation based trajectory data privacy protection method and system. According to the method, an individual privacy risk metric value is acquired through establishment of a privacy model, the individual privacy risk metric value and the actual condition of urban population risk spatial distribution are compared, privacy risk protection is established, and an effect after privacy protection is evaluated. The method is simple, low in calculation complexity and applicable to spatial data with large data volume.

The invention provides a system solution for protecting position privacy on the premise of strictly ensuring privacy, belonging to the field of data information security. For most spatial transformation techniques, the proposed method relies on semantic privacy model, such as k-anonymous, or ad hoc uncertainty models that do not provide strict privacy. Many of them consider only static scenarios or disturbance locations on a single timestamp, rather than the temporal dependency of the mobile user's location, and are therefore vulnerable to various inferential attacks to propose solutions. Firstly, according to the topological relation of the geographic space, the algorithm is proposed to calculate the privacy level of each region on the map. Secondly, a differential privacy based onsigma-location set is proposed to protect the real position of each time point, and the sensitivity package is used to capture the geometric meaning of two-dimensional sensitivity. Finally, a special position perturbation mechanism is proposed to implement the algorithm based on sigma-Differential privacy of the location set.

The invention discloses a location anonymous method for continuous queries. The method comprises the steps that when an anonymous query is received, each unexpired query to be anonymous in a query set to be anonymous is detected; when the unexpired queries to be anonymous meet a quality model, the queries to be anonymous are inserted into a candidate anonymous set; when the detection of the queries to be anonymous in the query set to be anonymous is finished, and the number of queries in the candidate anonymous set is larger than or equal to that of a preset anonymity degree, the candidate anonymous set forms an anonymity circle, and whether the anonymity circle meets a privacy model or not is judged; when the anonymity circle meets the privacy model, the center of the anonymity circle is adjusted, and the anonymity circle is returned. The invention further discloses a location anonymous device for the continuous queries and a privacy protection system. The privacy protection on the continuous queries is improved, and location service quality and anonymity efficiency are improved.

Systems and / or methods that mitigate overexposure of shared content and that facilitate improving security of content in a computing environment are provided. In particular, the systems and methods involves monitoring resource activities such as user interactions (between users and / or between users and resources), email activity, share space activity, shared content activity, distribution lists, contact lists, personnel or organization structures, and the like. Data can be collected as one or more resource activities are monitored and then analyzed to determine where, if any, changes have taken place. For example, if a user has been removed from a distribution list or a user has not accessed the shared content within a time period, a recommendation can be formulated to suggest an adjustment to the affected access rights. Furthermore, the systems and / or methods can be integrated with other applications such as an email client, whereby an out-of-office indicator can leave access rights as they are.

The invention provides a privacy model training method based on a small amount of public data and an electronic device. The privacy model training method comprises steps of obtaining N neural networkteacher models by using a plurality of trainings; respectively inputting a small amount of public data xi into the N neural network teacher models to obtain a statistical voting result of each publicdata xi for each label k; noise being added to all statistical voting results, and public data xi and corresponding labels meeting the differential privacy principle being obtained; optimizing an adversarial generative network through a large number of random noise vectors and a pre-trained discriminant neural network, and generating a large number of label-free data; and jointly training a student model through the public data xi meeting the differential privacy principle, the corresponding label and a large amount of unlabeled data pair pre-trained auto-encoders to obtain a private student model. According to the method, only a small amount of public data is needed to train one private student model, physical isolation and network isolation of sensitive data are realized, and a problem of low accuracy of the private student model is solved.

The invention discloses a blockchain privacy protection method, system and device. The method comprises the steps of receiving an established privacy model containing a transaction format and a privacy requirement of a target transaction; respectively generating a local encryption SDK (Software Development Kit) containing an encryption API (Application Program Interface), a verification contract containing a transaction verification API and an algorithm library containing algorithms required by on-chain transaction processing according to the transaction format and the privacy requirement, andlinking the verification contract and the algorithm library; calling an encryption API to encrypt the transaction plaintext of the target transaction to obtain a ciphertext transaction; and after theciphertext transaction is linked, calling a transaction verification API to verify the ciphertext transaction, and after the verification is legal, performing service processing on the ciphertext transaction based on the algorithm library and the defined service contract to obtain a service processing result. Therefore, a brand-new privacy protection scheme is provided, a user only needs to provide a privacy model, the system can automatically execute model analysis, transaction encryption, transaction verification, transaction processing and other work, and the learning cost is low.

The invention discloses a centralized privacy protection method for decision model release, and aims to overcome the defects of the existing privacy protection technology for issuing the decision model. According to the method, firstly, decision model construction and privacy modeling are carried out respectively according to an intelligent decision scene data model and privacy protection requirements, then, sensitivity is calculated based on a privacy model, finally, Laplace noise is added to model parameters of the decision model to realize privacy protection, and the decision model which can be directly published is obtained. Compared with a common privacy protection method oriented to simple data statistics, the invention has the advantages that privacy modeling is carried out according to the data model in the decision-making scene, the defect of lack of privacy protection normal forms is made up theoretically, privacy protection is achieved based on Laplace noise disturbance, and meanwhile the accuracy of the decision-making model is guaranteed.

The invention belongs to the technical field of information security, and relates to a sensitive label track data publishing method based on a graph differential privacy model. The method comprises the steps: clustering and generalizing tracks, performing graph-based time-space point differential privacy and publishing track data with privacy labels. In the first stage, firstly, all space-time points are divided into hot spot areas and abnormal values, and then the center positions of the hot spot areas are used for generalizing each specific position in the trajectory data; in the second stage, a noise map is established, hot spots, abnormal values and privacy tags are mapped to a directed weighted graph, different privacy budgets are set for different vertexes in order to control the size of the added noise and protect the availability of the data as much as possible, and then Laplace noise is added to achieve differential privacy; in a publishing stage, each record in the generalized trajectory data set is restored, the overhead point is selected, the trajectory data is generated according to the heuristic expression, anonymous data is obtained, and high availability of the datais ensured.

The invention belongs to the technical field of mobile edges, and discloses a mobile streaming media edge cooperative distribution device and method based on differential privacy, and the device comprises a data privacy module and a cooperative decision module. When a streaming media cooperative transmission period starts, resource state information of an edge server and request video clip information of a streaming media watching user is collected. fuzzification request information of a streaming media watching user is acquired. A geographic position privacy model is created. Cooperation information of a relay transmission user is acquired. A cooperation cost privacy model and a reward function decision model are established. A collaborative user matching model is created and a streamingmedia content distribution path is determined according to the reward function decision model. Streaming media content edge cooperative distribution is completed, and related information is updated. The problem that in the prior art, the privacy of sensitive data cannot be guaranteed through streaming media cooperative transmission is solved, leakage of user privacy data can be prevented, and thecost of streaming media cooperative distribution is reduced.

The invention discloses a multi-attribute data privacy removing method taking practicability into consideration. The method comprises the following steps of S1, importing preprocessed multi-attribute data; S2, defining necessary attributes and sensitive attributes according to attribute descriptions, setting pre-grouping rules of the necessary attributes, and defining the sequences of the necessary attributes according to the attribute characteristics; S3, establishing a privacy exposure risk tree, and taking the attribute sequences and the pre-grouping rules in the step S2 as hierarchical sequences of the privacy exposure risk tree and the basis for generating branches of each hierarchy; and S4, measuring result information according to coding risks of the sensitive attributes on nodes and borders of the privacy exposure risk tree. According to the method, the suitable method can be flexibly selected from a plurality of common grammar anonymization and difference privacy models for solving the privacy problem, thereby satisfying various privacy demands for different data. Through utilization of special advantages of the privacy exposure risk tree, a multi-dimensionally aggregated space is designed compactly.

The invention discloses a high-dimensional position trajectory data privacy protection publishing system and method based on TP-MFSA suppression publishing, and belongs to the field of privacy protection algorithms. In order to resist an identity link attack and an attribute link attack caused by a track sequence in a high-dimensional track data release process and consider the data set mining practicability after anonymous processing, the invention provides a high-dimensional position trajectory data privacy protection publishing system and method based on TP-MFSA inhibition publishing. The method comprises: finding a minimum violation sequence set with privacy leakage according to a KCL-privacy model; judging the suppression mode of each time-space point in the MVS, namely, local suppression or global suppression; and designing an inhibition sequence model of each time-space point in the MVS based on the privacy protection intensity and the data practicability, and performing inhibition operation on the data set according to the sequence to obtain a publishable high-dimensional trajectory data set.

The invention discloses a federated learning method based on one-time clustering and inter-cluster prototype alignment, and the method comprises the steps: firstly, each client calculates and uploads a feature signature based on a distributed isomorphic proxy model; and the server performs one-time clustering on the clients based on the feature signatures. In the subsequent iteration training, the server distributes the corresponding cluster level agent model and the global class prototype to each cluster client; each client maintains a heterogeneous privacy model and an isomorphic proxy model, and performs joint training; after training, the client uploads the updated proxy model and a newly generated local class prototype; and the server aggregates the proxy model in each cluster, aggregates the local class prototypes of all the clients to update the global class prototype, and continuously iterates until the model converges. By using the method, huge communication and calculation overhead caused by repeated clustering in the training process is avoided. The method can be widely applied to the technical field of distributed machine learning.

The invention provides a data cleaning system for a terminal agent, relates to the technical field of data processing, and realizes efficient and intelligent purification of original heterogeneous data through combination of a preliminary cleaning module and a multi-model collaborative risk detection module. Four artificial intelligence models of the risk detection module, namely a security model, a sensitive privacy model, a networking large model and a poisoning detection model, work cooperatively, so that the system can accurately identify complex risks such as network attacks, personal privacy leakage, false information and data poisoning, and automatic shunting and disposal are realized based on a risk grading mechanism. And the comprehensiveness, accuracy and intelligent level of data cleaning are improved.

The present invention discloses a system and method for secure information processing that guarantees data privacy, security, and auditability throughout the AI lifecycle. The system empowers the user to classify input data as "public" or "private," which dictates its routing into one of two segregated processing pathways. The public pathway handles non-sensitive information using external AI models, while the private pathway processes sensitive data in a secure, isolated environment. A novel federated embedding capability safely synthesizes insights from both pathways, enhancing the overall quality of the output through their combination, without exposing raw private data. Underpinning the architecture is a pervasive blockchain layer that provides an immutable, cryptographically verifiable record of every action, enforcing data handling policies and ensuring provable security and compliance for regulated industries.

The invention provides a privacy protection method and system based on a high-risk frequent active point replacement policy. The privacy protection method comprises the following steps: establishing a privacy model, grading individual risks into zero-risk individuals, low-risk individuals and high-risk individuals based on the privacy model, hiding the high-risk individuals in the low-risk individuals for privacy protection, and evaluating a privacy protection result. The method is simple, is low in calculation complexity and can be applicable to spatial data with large data sizes.

The invention discloses a vehicle position privacy protection method based on group signature and reputation, which comprises a reputation model for encouraging more vehicles to cooperate with each other to form a group, a privacy model for evaluating the position privacy strength of the vehicle, a group forming and dynamic mixed region establishing mechanism, and a pseudonym replacement mechanism. Any vehicle in the vehicle-mounted ad hoc network actively requests establishment of a group from a control server according to the reputation value of the vehicle when the pseudonym of the vehicleis soon to be expired; the control server allocates a group leader role to the requesting vehicle and dynamically forms a mixed region according to the current position of the requesting vehicle; theother vehicles in the mixed region judge whether the cooperation is participated in or not to change the pseudonym according to the reputation value, privacy strength and the remaining time of the pseudonym of the other vehicles; and the reputation value of the vehicles can be dynamically adjusted. According to the method, more vehicles can be encouraged to join the group, thereby changing the pseudonym in cooperation and ensuring the position privacy of the vehicles; meanwhile, the vehicles do not need to keep a silent state in the formed mixed region but broadcast messages as a group identity so that the running safety of the vehicles is guaranteed; furthermore, resource expenditure caused by unnecessary frequent change of the pseudonym can be effectively reduced.

The invention provides a privacy protection method and system based on a high-risk frequent active point replacement policy. The privacy protection method comprises the following steps: establishing a privacy model, grading individual risks into zero-risk individuals, low-risk individuals and high-risk individuals based on the privacy model, hiding the high-risk individuals in the low-risk individuals for privacy protection, and evaluating a privacy protection result. The method is simple, is low in calculation complexity and can be applicable to spatial data with large data sizes.

The invention relates to a privacy model training method and device, computer equipment and a storage medium. The method of one embodiment comprises the steps of issuing a model training task based on a to-be-trained model; receiving a training joining request sent by each user terminal, generating a user set according to the request time of the training joining request of each user terminal, and sending the user set to each user terminal; dividing the users in the user set into user groups; in any iteration training process, receiving a model training ciphertext which is sent by each user terminal in the user set and carries a user terminal ciphertext, a model parameter ciphertext and training iteration times; based on the user group to which each user terminal belongs, performing key aggregation processing on the user terminal ciphertext of each user terminal to obtain a key aggregation processing result; and when the key aggregation processing result passes verification, performing model aggregation processing on the model parameter ciphertext of each user terminal to obtain a model aggregation result. By adopting the method, the safety and the stability during model training can be improved.

The invention belongs to the technical field of information security, and provides a privacy measurement algorithm for structured data in a data release environment. Firstly, a data source collects user information and generates a user information abstract, an external data set is screened according to abstract information, and attribute similarity is calculated to distinguish private attributes and non-private attributes. Then, external and internal average information amount and distribution information amount of the private attributes are calculated, sensitive attributes for the private attributes are priced and identified based on demand and time loss, quasi-identifiers are extracted by applying a maximum entropy principle, and risk elements are identified in the data. Finally, the data privacy is measured according to the privacy model, a privacy-oriented anonymous technology is given, and it can be proved that the privacy and availability under the orientation reach the best balance.

The invention discloses an MEC task unloading method based on opportunity constraint, and relates to the technical field of cloud computing, and the method comprises the steps: building a time delay energy consumption model, a fuzzy privacy model and an opportunity constraint model through the fuzzy data input quantity and the fuzzy CPU cycle number of a fuzzy to-be-processed task; establishing a joint optimization model through a time delay energy consumption model, a fuzzy privacy model and an opportunity constraint model; solving the joint optimization model to obtain a task unloading strategy and a network selection strategy; distributing the fuzzy to-be-processed task to an unloading position according to a task unloading strategy; and performing resource transmission according to the network selection strategy. According to the method, the joint optimization model is solved through the designed unloading algorithm, and the optimal solution about the task unloading position strategy and the network data transmission selection strategy is obtained. And distributing the tasks to the corresponding unloading positions. And according to the network data transmission selection strategy, distributing task data to different networks according to the proportion so as to transmit the data.

The invention provides a cloud terminal data and platform sharing method and system based on differential privacy and a storage medium, and relates to the field of cloud data processing. The method comprises the steps of collecting and preprocessing operation data of a user terminal, sending the operation data to a cloud platform based on an operation state of the user terminal, and realizing complete collection of the user terminal to ensure richness of data sources of the cloud platform; screening available operation data from the preprocessed operation data received by the cloud platform and storing the available operation data in a storage space of the cloud platform; based on a data acquisition request of the trusted entity, a differential privacy optimizer is configured to train a differential privacy model, data expected to be acquired by the trusted entity is processed into privacy data, and the privacy data is sent to the trusted entity; according to the method, parameter adjustment is carried out on the differential privacy optimizer on the basis of the processing log of the trusted entity on the privacy data, and when the trusted entity requests to obtain the data again, the differential privacy optimizer is used for training the differential privacy model again, so that the credibility and the security of data sharing are improved.

The invention relates to the technical field of federated learning privacy defense, in particular to a privacy enhancement method for federated learning classification task gradient attacks, which comprises the following steps of: S1, initializing global model parameters, and performing batch processing division on participant training samples participating in federated learning; s2, a local micro-batch gradient is calculated, and the local micro-batch gradient is calculated; step S3, according to the micro-batch gradient calculated in the step S2, threshold limit cutting is carried out, and a gradient parameter is obtained; step S4, micro-batches are used as a sample unit for adding noise, gradient parameters of participants with strong privacy requirements are frozen, and Gaussian noise is added into the clipped gradient parameters; s5, gradient parameter loss is calculated; and S6, aggregating the gradient parameter loss of all participants in the step S5 through an aggregation algorithm to update global model parameters in the training, and calculating the joint differential privacy model accuracy and loss budget. According to the method, the privacy protection granularity and the model performance are well balanced.

The invention discloses a hierarchical privacy protection method in the release of hierarchical data. First, the sensitive attributes of the original hierarchical data are classified according to the actual sensitivity, and then the frequency parameter α is set according to the divided grades. i To limit the frequency of different levels of sensitive attributes in the equivalence class, and then use the clustering method to achieve a given privacy model. The present invention well solves the problem of privacy leakage caused by the proximity attack on the sensitive attributes of the existing hierarchical data publishing technology, improves the protection of hierarchical data privacy, and enhances the security of data; the present invention is for The protection of hierarchical data has certain practical significance and practical value.