Patents by Inventor Gopikrishnan VARADARAJULU
Gopikrishnan VARADARAJULU has filed for patents to protect the following inventions. This listing includes patent applications that are pending as well as patents that have already been granted by the United States Patent and Trademark Office (USPTO).
-
Patent number: 11410081Abstract: In a secure multi-party computation (sMPC) system, a super mask is constructed using a set of masks corresponding to a set of data contributors. Each data contributor uses a corresponding different mask to obfuscate the data of the data contributor. a first scaled masked data is formed by applying a first scale factor to first masked data of the first data contributor, the scale factor being computed specifically for the first data contributor from the super mask. A union is constructed of all scaled masked data from all data contributors, including the first scaled masked data. A machine learning (ML) model is trained using the union as training data, where the union continues to keep obfuscated the differently masked data from the different data contributors. The training produces a trained ML model usable in the sMPC with the set of data contributors.Type: GrantFiled: May 20, 2019Date of Patent: August 9, 2022Assignee: INTERNATIONAL BUSINESS MACHINES CORPORATIONInventors: Vaibhav Murlidhar Kulkarni, Rakhi S. Arora, Padmanabhan Krishnan, Gopikrishnan Varadarajulu
-
Patent number: 11288397Abstract: Textual masking for multiparty computation is provided. The method comprises receiving masked input data from a number of contributors, wherein the input data from each contributor has a unique contributor mask value. A unique analyst mask factor is received for each contributor, computed by an analyst as a difference between a uniform analyst mask value and the contributor mask value. An API call is received from the analyst to aggregate the input data from the contributors. The respective analyst mask factors are added to the input data from the contributors, and the data is aggregated and shuffled. Computational results received from the analyst based on the aggregated input data are published. In response to API calls from the contributors, the analyst mask factors are removed from the computational results, wherein computational results received by each contributor are masked only by the respective contributor mask value.Type: GrantFiled: September 3, 2019Date of Patent: March 29, 2022Assignee: International Business Machines CorporationInventors: Padmanabhan Krishnan, Vaibhav Murlidhar Kulkarni, Gopikrishnan Varadarajulu, Rakhi S. Arora, Samir Katti
-
Publication number: 20210064779Abstract: Textual masking for multiparty computation is provided. The method comprises receiving masked input data from a number of contributors, wherein the input data from each contributor has a unique contributor mask value. A unique analyst mask factor is received for each contributor, computed by an analyst as a difference between a uniform analyst mask value and the contributor mask value. An API call is received from the analyst to aggregate the input data from the contributors. The respective analyst mask factors are added to the input data from the contributors, and the data is aggregated and shuffled. Computational results received from the analyst based on the aggregated input data are published. In response to API calls from the contributors, the analyst mask factors are removed from the computational results, wherein computational results received by each contributor are masked only by the respective contributor mask value.Type: ApplicationFiled: September 3, 2019Publication date: March 4, 2021Inventors: Padmanabhan Krishnan, Vaibhav Murlidhar Kulkarni, Gopikrishnan Varadarajulu, Rakhi S. Arora, Samir Katti
-
Publication number: 20200372394Abstract: In a secure multi-party computation (sMPC) system, a super mask is constructed using a set of masks corresponding to a set of data contributors. Each data contributor uses a corresponding different mask to obfuscate the data of the data contributor. a first scaled masked data is formed by applying a first scale factor to first masked data of the first data contributor, the scale factor being computed specifically for the first data contributor from the super mask. A union is constructed of all scaled masked data from all data contributors, including the first scaled masked data. A machine learning (ML) model is trained using the union as training data, where the union continues to keep obfuscated the differently masked data from the different data contributors. The training produces a trained ML model usable in the sMPC with the set of data contributors.Type: ApplicationFiled: May 20, 2019Publication date: November 26, 2020Applicant: International Business Machines CorporationInventors: Vaibhav Murlidhar Kulkarni, Rakhi S. Arora, Padmanabhan Krishnan, Gopikrishnan Varadarajulu
-
Patent number: 9292572Abstract: Database tables can have different types of database indices defined for the database tables and different numbers of database indices. The efficiency of reading the indexes can vary with the different profiles of the indexes, which impacts the costs of access plans that use the indexes. Weights can be predefined to reflect the relative efficiencies of the different characteristics. Costs can be computed in accordance with a variety techniques (e.g., based on edge traversals). The weights can be predefined to reduce costs, increase costs, or a combination thereof. A database management application or associated application or program can also refine or revise these weights based on statistical data gathered about the operation of the database and/or heuristics that are developed based on observations/research. The corresponding weights can be adjusted accordingly.Type: GrantFiled: June 26, 2014Date of Patent: March 22, 2016Assignee: International Business Machines CorporationInventors: Abhinay R. Nagpal, Sandeep R. Patil, Gopikrishnan Varadarajulu
-
Publication number: 20140310261Abstract: Database tables can have different types of database indices defined for the database tables and different numbers of database indices. The efficiency of reading the indexes can vary with the different profiles of the indexes, which impacts the costs of access plans that use the indexes. Weights can be predefined to reflect the relative efficiencies of the different characteristics. Costs can be computed in accordance with a variety techniques (e.g., based on edge traversals). The weights can be predefined to reduce costs, increase costs, or a combination thereof. A database management application or associated application or program can also refine or revise these weights based on statistical data gathered about the operation of the database and/or heuristics that are developed based on observations/research. The corresponding weights can be adjusted accordingly.Type: ApplicationFiled: June 26, 2014Publication date: October 16, 2014Inventors: Abhinay R. Nagpal, Sandeep R. Patil, Gopikrishnan Varadarajulu
-
Patent number: 8856753Abstract: Processor performance in executing a section of software code is measured. A clock associated with the section of software code. It is detected that the section of software code is beginning to be executed. In response, the clock associated with the section of software code is started. It is then detected that the section of software code is no longer being executed. In response, the clock associated with the section of software code is stopped. The clock is stopped and restarted as the section of software code temporarily stops execution to cede control to other sections of software code. Ultimately, a value of the clock associated with the section of software code is output, where the value corresponds to the processor performance in executing the section of software code.Type: GrantFiled: April 18, 2008Date of Patent: October 7, 2014Assignee: International Business Machines CorporationInventor: Gopikrishnan Varadarajulu
-
Patent number: 8615507Abstract: There is disclosed a system and method for optimising database access plans. Specifically, modifications made to a database are monitored and compared to a set of criteria. If a modification satisfies a criterion, the access plans relating to the modification are updated.Type: GrantFiled: December 23, 2008Date of Patent: December 24, 2013Assignee: International Business Machines CorporationInventors: Gopikrishnan Varadarajulu, Abhishek Iyer, Pallavi Gupta
-
Publication number: 20120191701Abstract: Database tables can have different types of database indices defined for the database tables and different numbers of database indices. The efficiency of reading the indexes can vary with the different profiles of the indexes, which impacts the costs of access plans that use the indexes. Weights can be predefined to reflect the relative efficiencies of the different characteristics. Costs can be computed in accordance with a variety of techniques (e.g., based on edge traversals). The weights can be predefined to reduce costs, increase costs, or a combination thereof. A database management application or associated application or program can also refine or revise these weights based on statistical data gathered about the operation of the database and/or heuristics that are developed based on observations/research. The corresponding weights can be adjusted accordingly.Type: ApplicationFiled: January 26, 2011Publication date: July 26, 2012Applicant: International Business Machines CorporationInventors: Abhinay R. Nagpal, Sandeep R. Patil, Gopikrishnan Varadarajulu
-
Publication number: 20100161649Abstract: There is disclosed a system and method for optimising database access plans. Specifically, modifications made to a database are monitored and compared to a set of criteria. If a modification satisfies a criterion, the access plans relating to the modification are updated.Type: ApplicationFiled: December 23, 2008Publication date: June 24, 2010Applicant: International Business Machines CorporationInventors: Gopikrishnan Varadarajulu, Abhishek Iyer, Pallavi Gupta
-
Publication number: 20090265691Abstract: Processor performance in executing a section of software code is measured. A clock associated with the section of software code. It is detected that the section of software code is beginning to be executed. In response, the clock associated with the section of software code is started. It is then detected that the section of software code is no longer being executed. In response, the clock associated with the section of software code is stopped. The clock is stopped and restarted as the section of software code temporarily stops execution to cede control to other sections of software code. Ultimately, a value of the clock associated with the section of software code is output, where the value corresponds to the processor performance in executing the section of software code.Type: ApplicationFiled: April 18, 2008Publication date: October 22, 2009Applicant: INTERNATIONAL BUSINESS MACHINES CORPORATIONInventor: Gopikrishnan VARADARAJULU