利索-全球专利检索

  1. Login
  2. Sign Up

Search Results

113 records (took 0.038 seconds)

    Federated learning with dataset sketch commitment based malicious participant identification
    3.
    发明授权
    Federated learning with dataset sketch commitment based malicious participant identification 有权

    公开(公告)号：US12147879B2

    公开(公告)日：2024-11-19

    申请号：US17180972

    申请日：2021-02-22

    Applicant: International Business Machines Corporation

    Inventor： Lei Yu ,  Qi Zhang ,  Petr Novotny ,  Taesung Lee

    IPC: G06N20/00

    Abstract: Mechanisms for performing intelligent federated machine learning (ML) model updates are provided. A plurality of ML model updates, and a plurality of dataset sketch commitment data structures (sketches), are received from a plurality of participant computing systems. Each sketch provides statistical characteristics of a corresponding local dataset used by a corresponding participant to train a local ML model. A potentially malicious participant identification operation is performed based on an analysis of the plurality of sketches to identify one or more potentially malicious participants based on differences in sketches. ML model updates received from participant computing systems identified as potentially malicious participants are discarded to thereby generate a modified set of updates. The federated ML computer model is updated based on the modified set of updates.

    Decentralized asset identifiers for cross-blockchain networks
    5.
    发明授权
    Decentralized asset identifiers for cross-blockchain networks 有权

    公开(公告)号：US11968301B2

    公开(公告)日：2024-04-23

    申请号：US17016843

    申请日：2020-09-10

    Applicant: International Business Machines Corporation

    Inventor： Timothy Olson ,  Petr Novotny ,  Nitin Gaur

    IPC: H04L9/30 ,  H04L9/00 ,  H04L9/06 ,  H04L67/104

    CPC classification number: H04L9/30 ,  H04L9/0643 ,  H04L67/1044 ,  H04L9/50

    Abstract: Described are techniques for generating and employing decentralized asset identifiers for cross-blockchain network asset transfers, the techniques including registering a decentralized asset identifier to an asset with a local asset identifier, where the decentralized asset identifier is immutable. The techniques further include endorsing a proposed transaction for transferring the asset from a first controller associated with a first blockchain network to a second controller associated with a second blockchain network, where the proposed transaction utilizes the decentralized asset identifier. The techniques further include exchanging, in a document associated with the decentralized asset identifier and retrieved from an identity network, the first controller for the second controller.

    BLOCKCHAIN DATA BREACH SECURITY AND CYBERATTACK PREVENTION
    6.
    发明公开
    BLOCKCHAIN DATA BREACH SECURITY AND CYBERATTACK PREVENTION 审中-公开

    公开(公告)号：US20240048582A1

    公开(公告)日：2024-02-08

    申请号：US18489888

    申请日：2023-10-19

    Applicant: International Business Machines Corporation

    Inventor： Yacov Manevich ,  Artem Barger ,  Nitin Gaur ,  Petr Novotny

    IPC: H04L9/40 ,  H04L9/30 ,  G06N5/04 ,  G06N20/00

    CPC classification number: H04L63/1425 ,  H04L9/30 ,  G06N5/04 ,  G06N20/00 ,  H04L9/50

    Abstract: Systems, methods, and computer programming products leveraging the use of machine learning, cryptographic keys and blockchain technology for validating blockchain transactions. The disclosed systems, methods and products improve detection of malicious cyberattacks and fraud, while reducing occurrences of falsely invalidated transactions and improving overall blockchain security in both permissioned and permissionless blockchain networks. Classifiers are trained using machine learning and other classification techniques by building a transaction history to learn how to identify suspicious transactions on the blockchain. In permissionless and order-execute models of permissioned blockchains, cryptographic keys are publicly registered to guardians residing out of band, who may co-sign requests and override or resubmit transactions marked as suspicious by the classifiers. In an execute-order model of permissioned blockchains, one-time use keys may be registered with the certificate authority of the blockchain, and used to co-sign transactions that might appear suspicious, preventing false-positive identification of suspicious-looking transactions by the classifier.

Patent Agency Ranking