Microsoft Technology Licensing, LLC

Apple Inc.

Ryan M Rogers

Abhishek Bhowmick

Umesh Suresh Vaishampayan

Gaurav Kapoor

Andrew H Vyrros

Julien F Freudiger

Parvez Ahammad

Jing Wang

Subbu Subramaniam

Sean S Peng

Ryan T Tecco

John Duchi

David Anthony Durfee

Adrian Rivera Cardoso

Man Chun D Leung

Rahul Tandra

Vignesh Jagadeesh

Kartik R Venkatraman

Lin Xu

Sowmya Gopalan

Ya Xu

Manoj Rameshchandra Thakur

David Pardoe

Kalu Onuka Kalu

Guillaume Tartavel

Stephen Cosman

Alexis Hugo Louis Durocher

Jefferson Lai

Marcelo Lotif Araujo

Michael Chatzidakis

Thi Hai Van Do

Shawn F Ren

Bing Liu

Yajun Wang

Saket Kumar

Mark B Cesar

Devang N Jhaveri

Liangzhong Yin

Mingyang Hu

Sudhanshu Arora

Yu Chen

Vinyas Maddi

Shraddha Sahay

Gavin Castro Uathavikul

Neha Jain

Seunghyun Lee

Raymond E Ortigas

Humberto A Gonzalez Cabezas

Technologies for secure multi-party computation include computing first double-encrypted data, computing second double-encrypted data, and, in a trusted execution environment, executing a query on the first double-encrypted data and the second double encrypted data to create a query-processed double-encrypted data set. The trusted execution environment can provide the query-processed double-encrypted data set to a requester such as another computer, system, or process.

Secure multi-party computation

Embodiments of the disclosed technologies receive first event data associated with a first party application, receive second event data representing a click, in the first party application, on a link to a third party application, receive third event data from the third party application, convert the third event data to a label, map a compressed format of the labeled third event data to the first event data and the second event data to create multi-party attribution data, group multiple instances of the multi-party attribution data into a batch, add noise to the compressed format of the labeled third event data in the batch, and send the noisy batch to a second computing device. A debiasing algorithm can be applied to the noisy batch. The debiased noisy batch can be used to train at least one machine learning model.

On-device privatization of multi-party attribution data

Embodiments described herein provide a technique to crowdsource labeling of training data for a machine learning model while maintaining the privacy of the data provided by crowdsourcing participants. Client devices can be used to generate proposed labels for a unit of data to be used in a training dataset. One or more privacy mechanisms are used to protect user data when transmitting the data to a server. The server can aggregate the proposed labels and use the most frequently proposed labels for an element as the label for the element when generating training data for the machine learning model. The machine learning model is then trained using the crowdsourced labels to improve the accuracy of the model.

Distributed labeling for supervised learning

Technologies for digital content distribution include creating a data set that includes a plurality of entity data records each comprising a plurality of attribute values. By applying a first data security technique to the data set, a first query term and a matching subset of the plurality of entity data records are determined. A second data security technique is applied to the matching subset. When output of the second data security technique satisfies a first noisy threshold, the first query term is added to a set of query terms. A third data security technique is applied to the matching subset. When output of the third data security technique satisfies a second noisy threshold, the set of query terms is expanded to include a second query term. The set of query terms is used to perform the digital content distribution.

Building data-secure group profiles

Embodiments described herein provide for a non-transitory machine-readable medium storing instructions to cause one or more processors to receive, at a client device, a machine learning model from a server, detect a usage pattern for a content item, store an association between the content item and the detected usage pattern in local data, train the machine learning model using local data for the content item with the detected usage pattern to generate a trained machine learning model, generate an update for the machine learning model, privatize the update for the machine learning model, and transmit the privatized update for the machine learning model to the server.

User behavior model development with private federated learning

Embodiments described herein provide for a non-transitory machine-readable medium storing instructions to cause one or more processors to perform operations comprising receiving a machine learning model from a server at a client device, training the machine learning model using local data at the client device, generating an update for the machine learning model, the update including a weight vector that represents a difference between the received machine learning model and the trained machine learning model, privatizing the update for the machine learning model, and transmitting the privatized update for the machine learning model to the server.

Private federated learning with protection against reconstruction

Technologies for providing event-level data privacy for streaming post analytics data include, in some embodiments, receiving a data stream that includes instances of count data collected over a time interval, computing a true count breakdown that includes a set of sub-counts of non-public user interface interactions on the post, creating a noisy count breakdown by applying at least one differential privacy mechanism to the set of sub-counts, and streaming the noisy count breakdown instead of the true count breakdown to a computing device. At least one of the sub-counts is a count that is associated with a particular value of an attribute that has different possible values. The attribute is associated with the non-public user interface interactions on the post.

Event-level data privacy for streaming post analytics data

In an embodiment, the disclosed technologies include receiving a query that requests aggregate information about entity event data relating to digital content delivered digitally by an entity management system to entities of the entity management system, the query associated with a requester account; determining a first privacy allocation for the requester account; determining a first privacy value, the first privacy value computed based on the query and a selected privacy algorithm; deducting the first privacy value from the first privacy allocation to produce a first privacy balance; causing executing of the query on the entity event data and providing a result set in response to the query only if the first privacy balance indicates that the first privacy allocation has not been depleted.

Analytics interface with privacy management

Techniques for ensuring differential privacy in top-K selection are provided. In one technique, multiple items and multiple counts are identified in response to a query. For each count, which corresponds to a different item, a noise value is generated and added to the count to generate a noisy value, and the noisy value is added to a set of noisy values that is initially empty. A particular noise value is generated for a particular count and added to the particular count to generate a noisy threshold. The particular noise value is generated using a different technique than the technique used to generate each noise value in the set. Based on the noisy threshold, a subset of the noisy values is identified, where each noisy value in the subset is less than the noisy threshold. A response to the query is generated that excludes items that correspond to the subset.

Differentially private top-k selection

Embodiments described herein provide techniques to encode sequential data in a privacy preserving manner before the data is sent to a sequence learning server. The server can then determine aggregate trends within an overall set of users, without having any specific knowledge about the contributions of individual users. The server can be used to learn new words generated by user client devices in a crowdsourced manner while maintaining local differential privacy of client devices. The server can also learn other sequential data including typed, autocorrected, revised text sequences, sequences of application launches, sequences of purchases on an application store, or other sequences of activities that can be performed on an electronic device.

Privatized apriori algorithm for sequential data discovery

Growing community of inventors

Sunnyvale, CA, United States of America

Ryan M Rogers