公开(公告)号：US20250148788A1

公开(公告)日：2025-05-08

申请号：US18387461

申请日：2023-11-07

Applicant: Bank of America Corporation

Inventor： Dylan Bellinger ,  Sanjay Lohar ,  Kyle Mayers ,  Amado Martinez Suarez ,  Cody Searl ,  Rachel Howze

IPC: G06V20/40 ,  G06N20/00 ,  H04N21/454

Abstract: Various aspects of the disclosure relate to automated monitoring of content associated with an individual for indicators that the content may be a deepfake. A deepfake analysis training device trains artificial intelligence-based models based on validated content associated with one or more profiles. Each profile is associated with an individual. A deepfake analysis engine may be installed on computing devices to monitor access to content (e.g., multimedia content) for deepfakes associated with the one or more profiles, as configured. The deepfake analysis engine may be stand-alone application, an application extension, or may be integrated into third-party applications via application programming interface functions. Analyzed content is assigned a likelihood score that is compared to threshold values that indicates a likelihood that the content is a deepfake. When identified, deepfake content initiates an alert generation and/or blocking of the content otherwise the content is presented to a user.