55 results about "Privacy preserving" patented technology

A computer system implements a unified framework integrating an adaptive elastic funnel (AEF) with a convergent intelligence fabric (CIF) for flexible and contextualized multi-agent AI and human collaboration at scale. The system provides a universal multi-modal key-value subsystem for sharing partial computations across agents, implements a hybrid greedy / non-greedy placement strategy for dynamic memory management, orchestrates dynamic computational workflows and tensor workflows using hierarchical tensor-fragment scheduling, enables cross-agent orchestration with policy-based privacy preservation, and incorporates quantum-resistant secure memory enclaves. The architecture supports continuous learning without catastrophic forgetting, compositional reasoning across modalities, and secure task execution in distributed environments. This integration enables unprecedented computational efficiency, secure collaboration, and adaptive intelligence in high-dimensional decision-making environments while supporting incremental adoption through modular interfaces.

A method for privacy preservation for machine learning networks includes splitting a trained neural network into a first part and a second part. The first part is a privacy preservation (PP) encoder and the second part is a PP machine learning (ML) model. The method further includes retraining the PP encoder and the PP ML model. The method can improve artificial intelligence (AI) systems, optimize performance and support decision making in use cases including, but not limited to medical / healthcare, fraud detection, image recognition, predictive maintenance and connected vehicles.

Systems and methods disclosed herein perform privacy-preserving evaluations of artificial intelligence (AI) agents. A first AI agent associated with a first entity obtains a machine-readable data structure defining one or more operative boundaries for a second AI agent associated with a second entity. The system generates a unique fixed reference value representing the machine-readable data structure by applying a first transformation operation set, and transmits the unique fixed reference value to a multi-agent storage to store the value. The system receives, via the multi-agent storage, a verification artifact from the second AI agent that indicates an observed value based on internal operational data of the second AI agent corresponding to the operative boundaries. The first AI agent determines a verification status of the verification artifact by comparing the unique fixed reference value with the observed value, and autonomously generates a verification record including a representation of the verification status.

A novel approach is presented that enables human rescuers and human operators to explore and navigate treacherous and challenging environments safely and efficiently by providing an integrated, practical, flexible, and robust system. Leveraging advanced sensor deployment, adaptive communication networks, and real-time subject detection with AR-enabled navigation to enhance safety and real-time usability, a single practical solution is disclosed that is flexible, scalable, robust and error resilient, secure, energy-efficient, sustainable, and configured to optimize sensor placement and robot behavior to adapt to changes in the environment. Key innovations include federated learning for privacy-preserving model updates and multi-modal sensing for superior accuracy.

Systems and methods disclosed herein perform privacy-preserving evaluations of artificial intelligence (AI) agents. The system identifies an auditing AI agent from a set of auditing AI agents for assessing target AI agent sets. The system obtains a data structure that defines operative boundaries for a target AI agent set and generates a reference value by applying a first transformation operation set on the data structure. The system transmits the reference value to a multi-agent storage and receives, via the multi-agent storage, a verification artifact from the target AI agent set that indicates an observed value generated by applying a second transformation operation set on an artifact set generated by the target AI agent set. The system determines, via the auditing AI agent, a verification status and responsive to a particular artifact failing to satisfy one or more assessment metrics, generates an action set to modify the target AI agent set.

A privacy-preserving authentication method for anti-counterfeiting QR codes using Vision Transformer- (ViT-) based Federated Learning (FL) is provided. In the method, an individual client authenticates an anti-counterfeiting QR code as captured in an image presented to the individual client. A local machine-learning (ML) model of the individual client determines authenticity of the anti-counterfeiting QR code as captured in an image presented to the individual client. The local ML model is initialized as a pretrained ViT-based model, pretrained on the large-scale ImageNet dataset for processing an input image to determine authenticity of the anti-counterfeiting QR code as captured in the input image. The plurality of clients performs a cyclic weight transfer FL process to update respective local ML models of the plurality of clients according to instant pluralities of training data respectively owned by different clients in the plurality of clients while preserving training-data privacy among the different clients.

This invention relates to the field of data mining and behavior recognition technology, specifically to a privacy-preserving data mining and behavior recognition method based on machine learning. This invention assesses the gradient sensitivity of each network layer of a target machine learning model to obtain a privacy leakage risk score and determine the privacy protection level of each layer. Based on Rényi differential privacy, it dynamically allocates a privacy budget for each network layer according to the privacy protection level. A gradient masking mechanism is used to add noise matching the privacy budget to the gradients of each layer, preserving key gradients and completing model training. The behavior data to be recognized is then anonymized and input into the trained model to perform behavior recognition. This achieves differentiated privacy protection for each network layer and dynamic adaptation to the model training process, and is adaptable to various behavior recognition scenarios.

This application discloses a data processing method and system based on the fusion of multi-party secure computation and federated learning. The method includes: in response to a data fusion request, obtaining the gradient update features of the business dataset and the current federated learning strategy; generating a multi-party secure computation protocol based on the gradient update features, including node interaction order rules, privacy-preserving computation logic rules, and protocol update frequency rules; using the gradient data as protocol input, performing privacy-preserving fusion of the gradient data and the business dataset through lightweight encrypted interaction to obtain a fused encrypted dataset; extracting computation requirement sub-features and encryption requirement sub-features, and simultaneously optimizing the federated learning strategy and the multi-party secure computation protocol based on their correlation mapping, so that the optimized gradient computation process and the encryption processing process work together, realizing a deep fusion of federated learning and multi-party secure computation, ensuring data privacy while improving data processing efficiency.

This invention relates to the field of privacy-preserving machine learning, and particularly to a privacy-preserving data processing method and prediction system. The method includes: obtaining an encryption priority value obtained by weighted fusion of at least two complementary encryption priority quantification indices; dividing the original feature vector into a first feature subset and a second feature subset according to the encryption priority value; receiving plaintext data of the first feature subset and homomorphically encrypted data of the second feature subset from a data provider; inputting the plaintext data into a first neural network subnet to extract a first hidden layer representation; inputting the homomorphically encrypted data into a second neural network subnet; fusing the first and second hidden layer representations in the ciphertext domain, and outputting an encryption prediction result. This invention reduces computational overhead while protecting the privacy of sensitive features by quantifying encryption requirements in multiple dimensions and dynamically dividing features, combined with a hybrid neural network of plaintext and homomorphic encryption.

This invention discloses a privacy-preserving computer-aided diagnostic method based on inner product function encryption. The steps include: 1) A key generation center calls a group generation algorithm to generate a master key msk and public parameters pp; 2) Calculates and returns a key to the data owner based on msk, the data owner's identifier ID, and the input vector; 3) The model owner calculates D = e(h1, h2) based on pp, ID, and the input vector. r And then send the ciphertext to the computer-aided diagnostic device; 4) The data owner obtains the inner product operation result based on the ciphertext and the key, and then calculates the prediction result.

This invention relates to the technical field of digital marketing, and in particular to a method and system for optimizing advertising delivery strategies based on user profile analysis. The method includes the following steps: collecting multimodal behavioral data of users across multiple platforms, including text, images, audio, and video; constructing a dynamic user profile that integrates multimodal features; collaboratively training an advertising delivery strategy model using local user profile data from each platform based on a federated learning framework, and updating global parameters through secure aggregation; and dynamically generating and executing a personalized advertising delivery strategy, including creative selection, bidding, frequency, and timing, based on the trained model and real-time contextual information. This application achieves accuracy, adaptability, and cross-platform collaboration in advertising delivery through multimodal deep understanding, privacy-preserving federated learning, and reinforcement learning dynamic optimization, effectively improving delivery performance, user experience, and ROI, while ensuring data privacy and compliance.

The invention provides a non-invasive system and method for detecting hormonal levels through voice biomarkers to support women's brain health. The system captures voice samples, analyzes acoustic and semantic features, and integrates behavioral and physiological data to identify hormonal phases. An AI engine processes these inputs to deliver personalized, science-backed recommendations for cognitive resilience, emotional balance, and mental clarity. The platform supports daily engagement through guided voice journaling and integrates with wearable devices and health applications. Applications include early detection of hormone-related brain conditions, personalized lifestyle interventions, and large-scale research into hormonal influences on brain health. Privacy-preserving architectures ensure secure data handling while enabling longitudinal tracking and clinical integration.

A system and method for distributed data intersection using privacy preserving collaborative detection is disclosed The system includes a processor and a memory that causes the processor to verify data sets and configurations from a first and second entity interacting in a distributed data intersection to ensure data integrity and privacy during a data collaboration. The processor enables computations between the first and second entity using cryptographic protocols to identify common data points. The processor allows the first and second entity to retrieve the common data points pertaining to other entity using a oblivious data sharing technique while maintaining confidentiality of the data accessed during the interaction. The processor facilitates validation of data integrity and accuracy without revealing underlying information. The processor controls data-sharing policies ensuring that authorized data is disclosed based on a predefined criteria thereby enabling collaboration of data for re-identification while protecting privacy.

A method for authenticating a subject as an individual whose biometric data has been previously obtained uses asymmetrically loaded servers that are configured with one-time values for use in multi-party computing to perform the authentication. A first plurality of the servers stores sharded authentication data from the individual and a second plurality of the servers stores sharded enrollment data from the subject. A first subset of servers in the first plurality of servers and a second subset of servers in the second plurality of servers engage in the multi-party computing.

Methods, systems, and apparatus, including medium-encoded computer program products for selecting and displaying content in privacy preserving manners are described. A first trusted content platform (CP) that executes within a first trusted environment can receive a digital component (DC) response that can include: (i) data indicating constrained DCs and, for each constrained DC, distribution parameters; and (ii) data indicating at least one contextual DC selected according to contextual data. The first trusted CP can send to a second trusted CP a second DC request that can include set of constraining values and the distribution parameters. The first trusted CP can receive from the second trusted CP, data indicating (i) one or more constrained DCs selected and (ii) selection values generated based on the set of constraining values. The first trusted CP can select a DC based on a selection value for each contextual DC and the selection value.

A privacy-preserving authentication method for anti-counterfeiting QR codes using Vision Transformer- (ViT-)based Federated Learning (FL) is provided. In the method, an individual client authenticates an anti-counterfeiting QR code as captured in an image presented to the individual client. A local machine-learning (ML) model of the individual client determines authenticity of the anti-counterfeiting QR code as captured in an image presented to the individual client. The local ML model is initialized as a pretrained ViT-based model, pretrained on the large-scale ImageNet dataset for processing an input image to determine authenticity of the anti-counterfeiting QR code as captured in the input image. The plurality of clients performs a cyclic weight transfer FL process to update respective local ML models of the plurality of clients according to instant pluralities of training data respectively owned by different clients in the plurality of clients while preserving training-data privacy among the different clients.

This invention relates to the field of data security and privacy computing, specifically to a secret-sharing-based method for solving the k-th minimum value in a distributed dataset D, efficiently solving for the k-th minimum value under a semi-honest security model. The core of this method lies in adaptively managing the data state of the participants and designing and utilizing a lightweight secure comparison algorithm to determine each bit value of the k-th minimum value sequentially from the most significant bit to the least significant bit. While ensuring the correctness of the result, it reduces the number of interaction rounds and message volume, avoiding a linear increase in communication volume with the size of the participants and the data bit width, thus reducing the heavy reliance on multiple secure comparisons and bit-by-bit determinations. Simultaneously, it tightens the information leakage surface, allowing only the minimum necessary information that can be derived from the result to be output, avoiding the leakage of additional information.

This invention relates to a personalized learning recommendation method for e-commerce, belonging to the field of information recommendation technology. The method integrates user basic attributes and historical behavior data, deeply integrating multi-dimensional user profiles through adaptive weighting; it utilizes the association mining between products and knowledge points, establishing a product knowledge graph based on graph neural networks to achieve product topic clustering and multi-level knowledge point summarization; based on user profiles and product knowledge associations, it leverages deep reinforcement learning to dynamically plan personalized learning paths, intelligently recommending optimal resource sequences according to the user's current interests and knowledge level. The system collects user feedback information in real time and completes real-time adaptive adjustment of the recommendation strategy through local model updates and federated learning, achieving highly accurate and privacy-preserving personalized learning resource recommendations under interest shifts and behavioral changes.

This invention belongs to the interdisciplinary field of information technology, fintech, and distributed systems, specifically relating to a privacy-preserving multi-party collaborative reconciliation method and system based on value codes. More specifically, this invention relates to a trusted reconciliation scheme that utilizes programmable "value codes" as the core reconciliation carrier. It aims to solve problems such as inconsistencies in transaction records among multiple participants, cumbersome reconciliation processes, excessive manual intervention, difficulties in dispute resolution, and data privacy leaks by constructing a distributed, auditable, highly transparent, and privacy-preserving reconciliation infrastructure. This solution is applicable to various business scenarios requiring efficient, accurate, and trusted reconciliation, such as cross-institutional transaction clearing, supply chain collaborative reconciliation, multi-party profit sharing verification in platform-based economies, and cross-border business transaction reconciliation. This invention breaks through the limitations of traditional ledgers, inventing value codes as a measure of human value, and constructing a trusted reconciliation axiomatic system based on information economics and game theory. The core algorithm of the system integrates: a transaction energy level mapping operator based on log-normal distribution, a reconciliation utility evaluation model based on Milgrom's information value equation, and a non-cooperative game-theoretic collaborative engine based on Nash algorithm.

Methods, systems, and apparatus, including computer programs encoded on a computer storage medium, for generating dynamic digital content in privacy preserving ways are described. In one aspect, a method includes receiving, by a trusted server and from multiple content platforms, digital component data for digital components. The server received, from each content platform, dynamic content selection logic for selecting discrete content elements for digital components of the content platform. The server selects, from digital components for which digital component data is stored in a digital component repository, candidate digital components based at least on user data included in a digital component request. For each candidate digital component, the server executes the dynamic content selection logic of the content platform that provided the digital component data for the candidate digital component, the executing resulting in selection of a particular layout and a particular subset of content elements for the digital component.

This invention presents a vertical federated frequent itemset mining method and system with differential privacy protection, belonging to the fields of privacy-preserving computing and big data mining technology. Firstly, leveraging the hashing properties and merging capabilities of Flajolet-Martin sketches, this invention designs an intersection-friendly cardinality estimation algorithm, eliminating the error accumulation defects caused by the traditional "union-then-complement" approach. By combining Bayesian inference and independence model expectation correction, it directly estimates the intersection cardinality of cross-participant attribute combinations with high accuracy. A two-layer "filtering-pruning" mechanism is introduced, reducing the dimensionality of continuous attributes through attribute binning and constructing a pruning enumeration tree based on globally frequent items. The Apriori anti-monotonicity and dynamic min-heap maintenance strategy are used to quickly eliminate low-frequency and redundant candidate items, thereby compressing the search space. Under strict differential privacy constraints, this invention significantly improves the accuracy, sorting quality, and computational efficiency of frequent itemset mining compared to existing homomorphic encryption and traditional differential privacy schemes.

This invention provides an intelligent bed prediction and management method and system for dynamic scheduling of medical resources; belonging to the field of medical resource management and intelligent scheduling technology; the method operates in a parallel execution environment composed of a physical system and a digital twin. First, it aggregates cross-institutional data and constructs a real-time digital twin through a privacy-preserving computing framework; then, based on a dynamic heterogeneous graph, it uses spatiotemporal graph neural networks and causal inference to perform accurate demand prediction; when the prediction is unbalanced, a scheduling scheme is generated in the digital twin environment through multi-agent reinforcement learning and game negotiation; before the scheme is executed, it is verified by sand table simulation and stored on the blockchain to ensure credibility; finally, the online fine-tuning and continuous evolution of the prediction model, decision-making strategy and game mechanism are realized through execution feedback; this invention solves the problems of cross-institutional data barriers, inaccurate prediction, inefficient scheduling and unreliable execution, and improves the overall utilization efficiency and emergency response capability of regional medical bed resources.

This invention relates to the field of data processing technology and discloses a privacy-preserving data sharing method and system. The method includes: performing reversible privacy-preserving processing on raw data to obtain transformed data and restored mapping information; associating the transformed data with a data identifier of the data holder and uploading it to a data sharing platform, and associating the restored mapping information with the data identifier and storing it in a secure execution environment; authenticating the data access request from the data requesting end to generate a temporary access credential containing authorization policy information; submitting the temporary access credential to the secure execution environment and calling the access control contract in the secure execution environment to verify the temporary access credential; extracting the mapping subset corresponding to the authorization policy information from the restored mapping information; performing a reverse restoration operation on the transformed data to obtain target data, and returning the target data to the data requesting end. This invention can improve the efficiency of privacy-preserving data sharing.

This specification provides one or more embodiments of a large-scale model retrieval enhancement generation method and apparatus. First, a target index sequence obtained based on privacy-preserving retrieval is acquired, and the corresponding target dataset is determined from a local knowledge base and merged into multiple data sample subsets. Then, a word-by-word generation loop is executed. In each loop, a first-layer utility function is constructed based on the conditional probability distribution output by the text output model, and a first differential privacy perturbation is performed to obtain candidate words. Then, the relative probability advantage of the candidate words is determined, and a second-layer utility function is constructed, and a second differential privacy perturbation is performed to obtain target words. Simultaneously, the target words are appended to the context fragment to complete the update. After the termination condition is met, the target words determined in each loop are concatenated to obtain desensitized synthetic data, which is then sent to the data receiver for generating response results of the user query text through a large language model.

This application relates to the field of digital economic technology. It provides a digital economic statistics method, apparatus, device, and medium. The method includes: scheduling statistical task instructions via protocol to obtain a dynamic scheduling strategy; allocating a privacy budget to a set of sub-queries in a pre-defined indicator system based on the statistical task instructions to obtain a privacy budget allocation scheme; performing privacy-preserving collaborative computation on the local datasets of each participant based on the dynamic scheduling strategy and the privacy budget allocation scheme to obtain intermediate computation results; fusing the intermediate computation results to generate a target macroeconomic indicator; and generating a verifiable encrypted credential based on the target macroeconomic indicator and the data contribution records of each participant. This achieves the technical effects of improving the accuracy of privacy budget allocation, optimizing cross-domain collaborative computation efficiency, reducing indicator fusion bias, and enabling traceability of data contributions.

A system and method for unified consent-based access controlled data flow in digital supply chain is disclosed The system (100) includes a processor (105) and memory (110) that execute instructions to obtain user consent prior to data collection, capturing contextual parameters such as time, source, and purpose. The system records consent in a structured data repository (135) with uniquely indexed entries, generates distinct consent identifiers to represent user-specific preferences, and regulates data flow (365) in accordance with these identifiers. It evaluates privacy risks, identifies potential vulnerabilities (145), and applies protection controls (150) based on impact severity. Secure data transfer routes (155) and privacy-preserving techniques (160) such as encryption, anonymization, masking are implemented on consent based access control (180). The system maintains retrievable processing records, links user-initiated data-subject requests, verifies secure deletions, generates compliance proofs, and presents a unified integrated operational interface (170) for real-time monitoring of consent, safeguards, and compliance.

This application discloses a method and system for secure fusion query governance of multimodal biomedical data. The method includes: receiving a fusion query request; verifying its legitimacy through a blockchain smart contract; decomposing it into sub-query tasks based on a multimodal metadata ontology model and distributing them; each data holder node generating intermediate results identified by a unified pseudonym in its local privacy-preserving computing environment; performing privacy-preserving data alignment and aggregation operations; returning the fusion result and recording key events in the blockchain smart contract. This application achieves integrated governance that ensures data is available but not visible, queries are flexible, processes are auditable, and contributions are incentivized. It is applicable to the secure collaborative analysis of multimodal biomedical data such as genomics, imaging, and electronic medical records.

A projection smart contract includes a synchronized copy of a subset of data held in a private smart contract. The subset of data can include information required for an interaction with another private smart contract while omitting data from the first private smart contract that need not be shared for the transaction. The projection can also be permissioned to limit availability of even the subset of data to parties that require access to the subset of information. Thus, data can be shared between private smart contracts without disclosing private information within those smart contracts.

A system for privacy preserving artificial intelligence (AI) comprising: a first and second device interoperability system (DIS) coupled to an artificial intelligence analysis subsystem (AIAS). The first and second DIS receive a first and second plurality of data sets from coupled first and second pluralities of user devices respectively. The first and second DIS create first and second aggregated data sets based on the received first and second pluralities of data sets. The first and second DIS determine first and second subsets of model parameters for an AI model based on the first and second aggregated data sets; then transmit first and second sets of information based on the model parameter subsets to the AIAS. The AIAS creates a set of model parameters for the AI model based on the transmitted sets of information, and transmits the set of model parameters to the first and second DIS for deployment.