International Business Machines Corporation

Globalfoundries Inc.

Nili Guy

Gal A Shachor

Samuel Kallner

Yoav Rubin

Nir Naaman

Idan Ben-Harrush

Nir Rozenbaum

Ira Rosen

Lior Turgeman

Gil Sharon

Adar Amir

Omer Tripp

Avi Yaeli

Michael Masin

Asaf Adi

Dmitri Pikus

David Franklin Manning

Michael C Mansell

Ariel Landau

Andrei Kirshin

Alexander Dvinsky

Maya Barnea

Anat Berkman

Gilad Moredehi Saadoun

Shmuel Kalner

Selective locating of missing chat messages for an edge device is provided. The process includes sending a locate chat message request from a user's edge device to at least one other edge device of at least one other participant in a chat with the user via an end-to-end encryption messaging system. The locate chat message request identifies at least one search parameter for at least one missing chat message of the chat on the edge device. The process further includes, based on sending the locate chat message request, receiving at the edge device from the at least one other edge device, the at least one missing chat message of the chat identified via the at least one search parameter.

Edge device facility to selectively locate missing chat messages

Techniques for generating a set of Deep Learning (DL) models are described. An example method includes training an initial set of DL models using the training data, wherein a topology of each of the DL models is determined based on the parameters vector. The method also includes generating a set of estimate performance functions for each of the DL models in the initial set based on the set of edge-related metrics, and generating a plurality of objective functions based on the set of estimated performance functions. The method also includes generating a final DL model set based on the objective functions, receiving a user selection of a selected DL model from the final DL model set, and deploying the selected DL model to an edge device.

Adaptation of deep learning models to resource constrained edge devices

An example system includes a processor to train an artificial intelligence (AI) model on a training set of images on a client device associated with a particular user. Images in the training set of images are selected using face recognition from a digital camera roll on the client device based on a set of faces chosen by the particular user of the client device. All of images in the training set of images include a detected face from the set of chosen faces. The processor is to filter a group message received from a second device based on the trained AI model.

Filtering group messages

An example system includes a processor to monitor system resources and performance preferences. The processor is to select model fragments based on the system resources and the performance preferences. The processor is to also construct a running artificial intelligence (AI) model from the selected model fragments. The processor is to further automatically modify the running AI model using the model fragments in response to detecting a change in the system resources or a change in the performance preferences.

Modifying artificial intelligence models using model fragments

In an approach for a rule-based filtering system for securing password logins, a processor receives a password input on a user device. A processor determines whether the password requires filtering. Responsive to determining the password requires filtering, a processor filters characters from the password based on a set of filtering rules to create a filtered password. A processor determines whether the filtered password matches a stored password.

Rule-based filtering for securing password login

An example system includes a processor to receive an artificial intelligence (AI) model trained on a client device associated with a particular user. The processor is to filter a group message based on the AI model. The processor is to send the filtered group message to the client device.

A method of presenting hyperlink data. The method comprises identifying when a web browser running on a web browser client retrieves, in response to a web document data request submitted to a target server, a first web document data for displaying a first web document containing a hyperlink having a label for display and a target destination defining an address of a second web document, evaluating a risk from content of the second web document, generating by the web browser to a reference evaluation indication of the risk, and processing the web document data and the reference evaluation indication by the web browser for generating a presentation on the client terminal which combines the first web document data and the reference evaluation indication such that when the reference evaluation indication is presented when the label is presented by the web browser.

Hyperlink data presentation

According to some embodiments of the present invention, there is provided a method for revealing at least a portion of a disguised password on a display device. The method comprises receiving a plurality of password characters typed by a user, presenting a plurality of replacement symbols in a password field displayed by a display device of a terminal having a touch surface, each the replacement symbol being indicative of the reception of another of the plurality of password characters, detecting a fingertip which slides contiguously against the touch surface, and temporarily replacing the presentation of at least some of the plurality of replacement symbols with of the presentation of at least some of the plurality of password characters in a sequential manner correspondingly with a movement of the fingertip.

Password presentation management

A method of aiding report construction based on inference of implicit application level relationships is provided herein. The method may include the following steps: analyzing a model of a software application to yield data elements associated with base objects of the software application and relationships between the data elements; generating a list of reportable objects comprising the base objects and objects associated with the data elements and their relationships; ordering the reportable objects based on at least one of: content and usage of the reportable objects, to yield an ordered list; and presenting the ordered list in a form usable for building a report on the software application, wherein at least one of: the analyzing, the generating, the ordering, and the presenting, is carried out by at least one processor.

Aiding report construction based on inference of implicit application level relationships

A method of ad hoc trust delegation in human-based workflow systems using one-time access key is provided herein. The method may include the following steps: generating a unique access key to a workflow system in response to a delegation of trust to one or more users; associating the unique access key with a specified portion of a workflow within the workflow system; embedding the unique access key in a message sent to the one or more users; and providing the one or more users with a one-time access to the workflow system, such that the access is limited to the specified portion of the workflow, in response to applying the unique access key to the workflow system.

Growing community of inventors

Kiriat Ata, Israel

Nili Guy