Intuit, Inc.

Itsik Yizhak Mantin

Yaron Sheffer

Sapir Porat

Laetitia Kahn

Guy Shtar

Henry Venturelli

Ron Bitton

Keren Simchon

Gal Cohen

Idan Habler

Jonathan Rabin

Yael Mathov Gome

A method includes receiving, at a server from a user device, a user prompt segment to a large language model (LLM), obtaining an additional prompt segment from a prompt data source, identifying a electronic address in the prompt segment, replacing the electronic address with a placeholder to generate a updated prompt segment, generating a LLM prompt comprising the updated prompt segment and the user prompt segment, and sending the LLM prompt to the LLM. The method further includes receiving a response to the LLM prompt from the LLM, the response comprising the placeholder, replacing the placeholder with the electronic address to generate an updated response, and sending the updated response to the user device.

Malicious prompt management for large language models

At least one processor may receive at least one document and generate at least one respective summary of each respective at least one document, the generating comprising omitting information originally included in the at least one document that is irrelevant to a computing task. The at least one processor may generate at least one respective embedding of each at least one respective summary and a retrieval augmented generation (RAG) database comprising the at least one respective embedding. The RAG database may be deployed so that the RAG database is configured to provide RAG data to a large language model (LLM) responsive to prompts generated through the computing task.

Selective content-aware embeddings for targeted information retrieval

A method is provided for authenticating a user. A request to access a resource is received from a user agent. A cookie associated with the request is identified. The cookie includes a first subset of data that was previously used to authenticate the user. The cookie is validated based on the first subset of the data. Responsive to validating the cookie, a second subset of the data is retrieved from server-side storage. A risk decision is generated based on the first subset and the second subset. When the risk decision meets a threshold, the user is authenticated without presenting an authentication challenge, and access to the resources permitted.

Web-authorization using enhanced cookie

A computer-implemented method includes receiving data comprising a plurality of application programming interface (API) requests from a plurality of client devices. The method includes generating a plurality of API sessions based on the data, wherein each of the API sessions is associated with a corresponding client device of the plurality of client devices and includes a sequence of API requests originating from the corresponding client device. The method includes comparing each of the plurality of API sessions to one or more of a plurality of different patterns indicative of permissible API sessions determined based on training data. The method includes determining one or more API sessions of the plurality of API sessions generated based on the data are abnormal based, at least in part, on the comparing. Finally, the method includes performing one or more actions based on determining the one or more API sessions are abnormal.

Detection of abnormal application programming interface (API) sessions including a sequence of API requests

A computer-implemented method includes receiving training data that includes a plurality of API requests from a plurality of client devices. The method includes generating a plurality of permissible API sessions based on the training data. Each of the permissible API sessions is associated with a corresponding client device of the plurality of client devices and includes a sequence of API requests originating from the corresponding client device. The method includes applying a sequence embedding technique to the plurality of permissible API sessions to generate a plurality of embeddings and applying a dimensionality reduction technique to the plurality of embedding to generate a plurality of compact embeddings. The method includes storing each of the compact embeddings in a space partitioning data structure at storage locations within the space partitioning data structure that are determined based on similarities between the compact embeddings.

Detection of abnormal application programming interface (API) sessions including a sequence of API requests using space partitioning data structures

A computer-implemented method includes receiving training data including a plurality of API requests from a plurality of client devices. The method includes generating a plurality of permissible API sessions based on the training data. The method includes applying a sequence embedding technique to the plurality of permissible API sessions to generate a plurality of embeddings. The method includes applying a dimensionality reduction technique to the plurality of embeddings to generate a plurality of compact embeddings. The method includes applying a clustering technique to the plurality of compact embeddings to determine a plurality of different clusters of the compact embeddings. The method includes generating a plurality of patterns based on the plurality of different clusters. Each of the plurality of patterns is descriptive of permissible API sessions associated with a corresponding cluster of the plurality of different clusters.

The present disclosure provides techniques for recommending vendors using machine learning models. One example method includes receiving transaction data indicative of a transaction, identifying, based on the transaction data, an associated merchant of one or more merchants and an associated time window of one or more time windows, receiving historical transaction data indicative of one or more historical transactions, wherein the one or more historical transactions are associated with the associated merchant and the associated time window, generating features based on the transaction data and the historical transaction data, and determining, using a classification model, that the transaction is fraudulent, based on the generated features.

Growing community of inventors

Shoham, Israel

Itsik Yizhak Mantin