Cyberark Software Ltd.

Niv Rabin

Michael Balber

Asaf Hecht

Mark Cherp

Eran Shimony

Eli Shemesh

Noa Moyal

Gal Naor

Systems, methods, and apparatuses are disclosed for detection and prevention of adversarial attacks against large language models. Techniques may include receiving an input associated with a target large language model, analyzing the input with a pre-trained classification algorithm to determine a first deconstruction process to be applied to the input, and modifying the input with a first deconstruction model using the determined first deconstruction process. Techniques may also include determining a score of a likelihood of the input being adversarial based on an output of the first deconstruction model and by applying a classification model and updating at least one of the first deconstruction model or the classification model based on the score.

Detection and prevention of adversarial attacks against large language models

Disclosed embodiments relate to systems and methods for dynamically performing entity-specific security assessments for entities of virtualized network environments. Techniques include identifying an entity associated with a virtualized network environment, identifying a plurality of security factors, determining entity-specific weights to the plurality of security factors, and generating a composite exposure assessment for the entity. Further techniques include selecting at least two security factors of the plurality of security factors, identifying the weights corresponding to the selected security factors, and calculating the composite exposure assessment using the selected security factors and corresponding weights, analyzing the composite exposure assessment, and generating at least one of: a security recommendation based on the analysis to alter a scope of privileges of the entity, a notification providing an indication of the composite exposure assessment, or a visual representation of the composite exposure assessment of the entity.

Dynamically generating multi-factor entity risk assessments within virtualized environments

Disclosed embodiments relate to systems and methods for securely performing actions on a resource. Techniques include receiving a request by the entity to perform a privileged action on a resource, the request including a token associated with the entity; providing a first indication of the request to a first handler; providing a second indication of the request to a second handler configured to perform the privileged action on the resource, wherein when the privileged action includes a query, the second indication of the request is provided to a query handler, and when the privileged action includes a write command, the second indication of the request is provided to a command handler.

Isolation and authorization for segregated command and query database resource access

Disclosed embodiments relate to systems and methods for reducing lag time for progressive consumption of data content. Techniques include receiving an indication of requested data, the indication comprising: a data chunk size, and a number of data chunks, accessing a data cache, and performing a fetching operation comprising at least one of: if the data cache is empty, obtaining a first portion of the requested data from a database, or if the data cache is not empty, determining whether at least the first portion of the requested data is available in the data cache. Further techniques include providing for consumption the first portion of the data, identifying that a threshold has been reached, receiving updated values for the data chunk size and the number of data chunks, performing the fetching operation again based on the updated values, and providing for consumption a second portion of the requested data.

Growing community of inventors

Petach-Tikva, Israel

Niv Rabin