Abstract: One or more embodiments perform a set of digital signature operations in a secure element (SE) platform runtime environment executing on a SE processor of a SE hardware device. A system initializes a signature generation object in an SE platform runtime environment. The system determines, via the signature generation object, a private key corresponding to a hash-based signature protocol. The system generates, via the signature generation object, a digital signature of a message digest by utilizing the private key to execute the hash-based signature protocol on the message digest. The system outputs the digital signature to a hardware device.

Abstract: The present disclosure relates to machine learning (ML) models, and more particularly to novel techniques for computing performance metrics for Multioutput-Multilabel ML models. Novel techniques are described for computing the performance metrics in a parallel and distributed without having to store the entire dataset for which metrics are to be computed in the memory of a data processing system. Novel data structures are provided for performing the computations.

Abstract: A system performs a set of cryptographic operations at least by utilizing an API to cause execution of a set of one or more secure element (SE) applications within the SE platform runtime environment of a first computing entity. The set of cryptographic operations include generating a first shared secret, generating a ciphertext at least by encapsulating the first shared secret with a first public key associated with a second computing entity in accordance with an encapsulation algorithm, and transmitting the ciphertext from the first computing entity to the second computing entity. The second computing entity derives the first shared secret by decapsulating the ciphertext with a private key corresponding to the first public key. The first computing entity and the second computing entity then exchange at least one encrypted message, encrypted with an encryption key that includes, or is based at least in part on, the first shared secret.

Abstract: Machine learning techniques directed to span prediction for textual data are disclosed. As used herein, span prediction is the process of predicting the possible spans of text that can be assigned to a given entity type of a set of predefined entity types. To this end, a machine learning model can be trained to generate values that indicate the predicted probability that a given span of an identified set of spans within text of interest is appropriate for association with a given entity type of the set of predefined entity types. The predicted probability values may be used to determine whether a given span or spans is associated with a given entity type. The predicted spans can also be scored in some examples.

Abstract: The present disclosure relates to a system and techniques for resolving dangling references resulting from a dependency relationship between computing resource objects uncovered during a harvesting process. The techniques include, adding a computing resource object from a catalog of computing resource objects to a computing resource collection for a client and identifying one or more dependencies for the computing resource object. The techniques further include determining at least one unresolved dependency from the one or more dependencies, the at least one unresolved dependency including a second dependency on a second computing resource object outside of the computing resource collection. The techniques further include resolving the at least one unresolved dependency after the second computing resource object associated with the unresolved dependency has been added to the computing resource collection.

Abstract: One or more embodiments include operations associated with semantic classification of data columns. The operations may include receiving a set of data elements corresponding to a data column to be semantically classified, applying a machine learning model to the set of data elements to predict a set of candidate semantic types for the set of data elements, selecting a particular semantic type from the set of candidate semantic types based at least in part on a semantic fit score corresponding to the particular semantic type predicted by the machine learning model, and presenting the particular semantic type as a recommended semantic classification for the data column.

Abstract: Techniques are described herein for performing thread-local garbage collection. The techniques include automatic profiling and separation of private and shared objects, allowing for efficient reclamation of memory local to threads. In some embodiments, threads are assigned speculatively-private heaps within memory. Unless there is a prior indication that an allocation site yields shared objects, then a garbage collection system may assume and operate as if such allocations are private until proven otherwise. Object allocations in a private heap may violate the speculative state of the heap when reachable outside of the thread. When violations to the speculative state are detected, an indication may be generated to notify the garbage collection system, which may prevent thread-local memory reclamation operations until the speculative state is restored.

Abstract: A method includes receiving a training set of utterances for training a machine-learning model to identify one or more intents for one or more utterances, and augmenting the training set of utterances with out-of-domain (OOD) examples. The augmenting includes: generating a data set of OOD examples, filtering out OOD examples from the data set of OOD examples, determining a difficulty value for each OOD example remaining within the filtered data set of the OOD examples, and generating augmented batches of utterances including utterances from the training set of utterances and utterances from the filtered data set of the OOD based on the difficulty value for each OOD. Thereafter, the machine-learning model is trained using the augmented batches of utterances in accordance with a curriculum training protocol.

Abstract: Systems, methods, and other embodiments associated with defining an adaptive model for conversational interaction are described herein. In one embodiment, a computer-implemented method includes semantically processing a conversational input to separate the input into a plurality of structural entities. The method also includes constructing a query of a data provider based on one or more of the structural entities and a query paradigm. The method also includes transmitting the query to the data provider for execution. The method further includes presenting the query results through a user interface.

Abstract: A method for generating a query filter list includes obtaining set of training queries, each training query comprising a predicate and one or more accessed columns returned from evaluating the predicate, and transforming the set of training queries into a structure. The structure relates, for an accessed column and a training query, the predicate and a correlation value to the accessed column. The method further includes normalizing the structure into a normalized structure. The normalized structure grouping entries in the structure according to accessed column. The method further includes generating a generalized query from the normalized structure, and adding the generalized query to the query filter list.

Abstract: Techniques are described herein for probabilistic monitoring of high-frequency, low-latency database queries. In some embodiments, a probabilistic query monitoring system periodically samples active database sessions. For example, the system may generate sample data every one second or at some other sampling rate for each database session that is currently active. The sample data may include a mapping between query identifiers to sample counter values that are extracted at different sample intervals. The system may then estimate performance metrics for the set of active database based on the counter values sampled across consecutive sample intervals.

Abstract: A computer system may receive one or more requests for access to one or more cloud services and may store the one or more requests in a request log. The computer system may receive one or more access rules applicable to cloud service access rights. The computer system may aggregate the one or more requests of the request log to determine access requirements for a container, the container being configured to store one or more applications. The computer system may generate and store container access policies that define access of a container and the one or more cloud services, the container access policies based at least in part on the aggregated one or more requests and the one or more access rules. The computer system may send the container access policies to a request forwarder of a compute instance in a production environment.

Abstract: Embodiments predict an occurrence of one or more hot sockets among a meter installation of a plurality of smart meters. Embodiments receive historical data over a predefined time period from the plurality of smart meters, the historical data including, for each smart meter, an amount of daily consumption of electricity for the meter, historical meter events for the meter, and all service orders associated with the meter. Embodiments pre-process the historical data to generate hot socket features. Embodiments train a machine algorithm using the hot socket features, and use the trained machine algorithm to predict the occurrence of one or more hot sockets.

Abstract: Techniques for extensions of graphical user interfaces (GUIs) are disclosed. The system executes an application that displays a GUI. The system selects one or more interface elements for displaying within the GUI at runtime while executing the application. The system identifies a primary data type corresponding to content that is to be displayed or currently being displayed by the GUI. The system determines that the primary data type is mapped to a first target data type. Responsive to determining that the primary data type is mapped to the first target data type, the system identifies a first function associated with the first target data type. The system generates a first interface element for initiating execution of the first function associated with the first target data type. The system displays the first interface element concurrently with a display of the content within the GUI.

Abstract: A sampling approach for time-window based multi-stage sampling. The sampling approach can determine whether received communications are of a stratum that is rare and determine a sampling mechanism for the communication based on whether the stratum is rare. The sampling system defines multiple time windows for sampling communications received by a computing system. The time windows are segmented into multiple time intervals. A portion of the multiple time intervals are randomly selected for sampling. A portion of the communications received during the selected time intervals are captured for security assurance purposes.

Abstract: Knowledge graph guide and entity controlled techniques for generating synthetic doctor-patient conversations.

Abstract: Techniques for implementing and enforcing a security policy in a secure element are disclosed. The secure element enforces the security policy to grant and/or deny access, such as from an application processor, to configuration of the device peripheral components and access to data of the device peripheral components across one or more bus architectures, such as an I3C bus. Implementing an access control policy in a secure element allows execution of code within the isolated secure element hardware processor, preventing software attacks that may emanate from code running in the application processor. This design also benefits from hardware protections against physical attacks.

Abstract: A build system is disclosed that identifies the inputs used by a build process for securely building and deploying a piece of software to production. The build system comprises a build container and a build proxy server. The build container receives a set of initial inputs for performing a build and generates a build output (e.g., a target artifact) as a consequence of performing the build. The build proxy server monitors both internal interactions as well as external interactions (e.g., input dependency fetches from external artifact repositories) of the build container within and outside a network boundary defined around the build container. Based on the monitored interactions, the build proxy server identifies all the additional input components and/or input component dependencies used by the build container for successfully performing the build. The build container uses the identified components to perform the build and generate a target artifact.

Abstract: Operations of a system may include executing a provisioning process that includes provisioning a network entity with a digital certificate for use in a stateless validation protocol. After provisioning the network entity with the digital certificate, the system may include receive a credential request from the network entity that includes the digital certificate and a request for an access credential for accessing a cloud resource. In response to the credential request, the system may execute an access-authorization process with respect to the network entity, including authenticating the digital certificate in accordance with the stateless validation protocol. Upon determining that the network entity authorized to receive an access credential, the system may provision the network entity with the access credential. The network entity may then use the access credential to access the cloud resource.

Abstract: A distributed event processing system is disclosed that receives a batch of events via a continuous data stream and performs the serialization of data in the batch of events. In certain embodiments, the system identifies a first data type of a first attribute for each event in a batch of events and determines a first type of data compression to be performed on data values represented by the first attribute. The system determines a first type of data compression to be performed on data values represented by the first attribute based on the first data type of the first attribute. The system then generates a first set of serialized data values for the first attribute. The system processes the first set of serialized data values against a set of one or more continuous queries to generate a first set of output events.