Microsoft Technology Licensing, LLC

Haitao Wang

Suman Kumar Nath

Nathan Jeffrey Ide

Sean Anderson Bowles

Chetan Bansal

Jack Smith Richins

Fabricio Voznika

Robert Bradley Gilbert

Ibrahim Mohammad Ismail

Paul J Miller

Jasraj Dange

Joseph B Isenhour

Manish Shetty Molahalli

Sagar Bholanath Saha

Steven James Syfuhs

Andrew James Wiley

Matthew Jonathan Joplin

Mitchell O Lindgren

Kamil Ozgur Arman

Siamak Ahari

Sergeii Gubenko

Machine-learned prediction of a blame frame of a crash stack. Specifically, a crash stack associated with a crash is parsed into a sequence of frames. The blame frame of the crash stack is estimated by, for each of a plurality of the sequence of frames, identifying a plurality of features of the corresponding frame, feeding the plurality of features to a neural network, and using the output of the neural network to make a prediction on whether the corresponding frame is a blame frame of the crash. If this is done during training time, the predicted blame frame can be compared against the actual blame frame, resulting in an adjustment of the neural network. Through appropriate featurization of the frames, and by use of the neural network, the prediction can be made cross-application and considering the context of the frame within the crash stack.

Crash localization using crash frame sequence labelling

An authentication server may not support all types of user credentials. For example, an on-premise authentication server may support credentials based on user secrets (i.e. username and password) and certificate-based credentials, but not hardware-key based credentials. A client device may use an un-supported type of credential to access resources managed by the on-premise authentication server by authenticating with a web-based authentication server. The web-based authentication server may support any type of credential, and the supported types of credentials may change over time. The web-based authentication server returns an authenticated user token indicating the user has been authenticated, but without authorizing access to any resources. The client device uses the on-premise authentication server to exchange the authenticated user token for an authorized user token. The client device then uses the authorized user token to access resources on the on-premise network.

Remote authentication for accessing on-premises network devices

Methods and devices for using candidate accounts on a computer device may include creating on the computer device, a candidate account representing a blank user profile for use on the computer device. The methods and devices may include automatically logging in the candidate account and locking the computer device. The methods and devices may include creating a user specific isolated environment on the computer device for the candidate account.

Candidate user profiles for fast, isolated operating system use

Resource use is recorded with a partitioned reference counter. The sum of all resource counter partitions is equivalent to the total references to a resource. When one resource counter partition reaches zero, it is possible that the resource should be destroyed. To determine if this is the case, all other partitions can be checked for a value of zero. If all the partitions are zero, the resource can be destroyed. Coarse grained partitioning and add/release on all partitions can be employed to avoid extra work associated with a local partition reaching zero. Further, destroying or deleting a resource can be accomplished in a manner that avoids a race condition.

Partitioned reference counter

Growing community of inventors

Redmond, WA, United States of America

Haitao Wang