International Business Machines Corporation

Dhinakaran Vinayagamurthy

Pankaj Satyanarayan Dayama

Vinayaka D Pandit

Sandeep Nishad

Santosh Ravi Kiran Penubothula

Krishnasuri Narayanam

Venkatraman Ramakrishna

Akshar Kaul

Nitin Singh

Angelo De Caro

Ermyas Abebe

Saneem Ahmed Chemmengath

Abhishek Singh

Dayama Pankaj Satyanarayan

Utsav Singhal

Harsh Chaudhari

Yining Hu

A system and method for a multi-party computation (MPC) is provided. In implementations, a method includes identifying a blockchain network and a computing device to perform an MPC based on an index; generating an MPC request including a function to be performed by the blockchain network and the computing device, data required for the function, and a verification policy defining a verification protocol to be performed by the blockchain network and the computing device; sending the MPC request to the blockchain network and the computing device; and receiving responses from a representative computing node of the blockchain network and the computing device, wherein each of the responses includes: an output of an MPC protocol performed by the blockchain network and the computing device to jointly compute the function while keeping the inputs private from one another and private from the computing system; and a proof based on the verification policy.

Collaborative computation across blockchain networks

A computer-implemented method for automatically and digitally securing an asset is disclosed. The computer-implemented method includes hashing a secret preimage to produce a secret hash. The computer-implemented method further includes locking a digital collateral asset with the secret hash for a first predetermined period of time. The computer-implemented method further includes locking a digital loan asset with the secret hash for a second predetermined period of time, wherein the second predetermined period of time is less than the first predetermined period of time. The computer-implemented method further pledging a digital loan repayment up to the amount of time of the first predetermined period of time plus the second predetermined period of time. The computer-implemented method further includes responsive to receiving a pledge for the digital loan repayment, automatically unlocking the digital collateral asset for a borrower of the digital loan asset.

Method for digitally securing an asset

A group of transactions to be rolled back in a blockchain ledger can be identified. Whether the transactions in the group are eligible to be rolled back together can be determined. Other transactions can be recommended, and the other transactions may be added to the group to be rolled back together. A rollback transaction can be added to the blockchain ledger that rolls back effect of the transactions, where the rollback transaction has zero or more inputs and zero or more output, the rollback transaction having at least one selected from at least one input and at least one output. An unspent output of a transaction in the group is provided as an input of the rollback transaction. Input of a transaction in the group, that is not an output of another transaction in the group, is provided as an output of the rollback transaction.

Blockchain ledger rollback

One embodiment provides a method, including: receiving, from a user at a collaboration platform, a request to perform a computation; generating a workflow comprising a sequence of steps for performing the computation; identifying potential data sources comprising the type of data and able to assist in performing at least one of the sequence of steps of the workflow; selecting computation data sources that collaborate to perform the computation, wherein the selecting is performed dynamically and based upon characteristics of a network created by the collaboration platform and between the computation data sources; and facilitating performance of the computation by the computation data sources using data of the computation data sources, wherein during performance of the computation the computation data sources collaborate within the network to perform the workflow while maintaining individual privacy of the data of the computation data sources and providing proof verifying a trustworthiness of the computation.

Platform for dynamic collaborative computation with confidentiality and verifiability

Methods, systems, and computer program products for a secure database-as-a-service are provided herein. A computer-implemented method includes obtaining, by a database service, data associated with one or more client devices to be stored in at least one encrypted database; encrypting, without mutation and in accordance with one or more security requirements, at least a portion of the data using an order preserving encryption scheme, wherein the encrypting comprises (i) computing a set of encryption parameters for the portion of the data and (ii) applying a process that converts a fixed-length cipher text corresponding to the portion of the data to a variable-length cipher text; and storing the encrypted data in the at least one encrypted database, wherein the database service enables one or more indexes to be built over the encrypted data to improve performance of query processing.

Secure database-as-a-service system

One embodiment provides a method, including: receiving a query from a user; providing the query to data owners, wherein each of the data owners has a local machine learning model and wherein the plurality of data owners train a meta-model; secret sharing model output from the data owners between the other data owners, wherein the model output comprises an output responsive to the query computed using the local machine learning model; receiving, from each of the plurality of data owners, a set of meta-features corresponding to the query; and generating a response to the query, wherein the generating comprises determining, by evaluating the meta-model using the set of meta-features received from each of the plurality of data owners, weights for outputs from the local machine learning models and aggregating the outputs in view of the weights.

Secure ensemble training and inference using heterogeneous private machine learning models

One embodiment provides a method, including: receiving, at a database proxy acting as an intermediary between a plurality of database clients and a service provider providing data management services for the plurality of database clients, a set of queries, of at least one of the plurality of database clients, for data stored at the service provider in an encrypted form, wherein the database proxy maintains a security budget defining a maximum threshold amount of data leakage for the plurality of database clients; batching the set of queries into query batches; transforming, for each query batch, each query within the query batch, wherein the transforming includes changing the query to reduce data leakage; performing, responsive to transforming each query within the query batch, a transformation on each of the query batches to reduce data leakage; executing, at the database proxy and utilizing an order-preserving encryption algorithm, the query batches; and calculating a remaining security budget based upon data leakage resulting from the executing.

Protection of data of database clients from persistent adversaries

One embodiment provides a method, including: receiving, at a service provider and from a user, a request for computation of a function, wherein the request includes a description of the function to be computed and requirements of the user; identifying, from data sources accessible to the service provider, a plurality of data sources to participate in the computation in view of the requirements and the description of the function, wherein each of the plurality of data sources includes data of a data owner and constraints on use of the data; identifying a multi-party computation protocol and parameters of the multi-party computation protocol for performing the computation; and running the computation of the function using the data of the plurality of data sources and in view of the requirements of the user and the constraints of the plurality of data sources.

Automation of multi-party computation as a service according to user regulations and requirements

Methods, systems, and computer program products for building an encrypted document store are provided herein. A computer-implemented method includes building an encrypted ordered key-value store; encrypting one or more sensitive keys to be stored in the ordered key-value store using an order preserving encryption scheme; storing a state of the order preserving encryption scheme in the ordered key-value store; and encrypting one or more values corresponding to the one or more sensitive keys using a semantically secure encryption scheme.

Building an encrypted document store

One embodiment provides a computer implemented method, including: receiving, at a service provider from each of a plurality of data owners each having data accessible to the service provider, a privacy budget, wherein the privacy budget identifies a set of privacy requirements to be employed by the service provider on data of the data owner; receiving, at the service provider, a query from a client; receiving, at the service provider from each of at least a subset of the data owners, a response to the query; adding, by the service provider, noise to each of the responses, wherein an amount of noise added to each response is based upon the privacy budget of the data owner corresponding to a given response; and returning, by the service provider to the client, an aggregated response including the responses having added noise.

Verifiable service provider performance regarding query response provision

Methods, systems, and computer program products for verifying confidential machine learning models are provided herein. A computer-implemented method includes obtaining (i) a set of training data and (ii) a request, from a requestor, for a machine learning model, wherein the request is accompanied by at least a set of test data; obtaining a commitment from a provider in response to the request, the commitment comprising a special hash corresponding to parameters of a candidate machine learning model trained on the set of training data; revealing the set of test data to the requestor; obtaining, from the requestor, (i) a claim of performance of the candidate machine learning model for the test data and (ii) a proof of the performance of the candidate machine learning model; and verifying the claimed performance for the requestor based on (i) the special hash and (ii) the proof of the claimed performance.

Growing community of inventors

Bengaluru, India

Dhinakaran Vinayagamurthy