Authenticity assessment of modified content

Number: 10951958
Published: 2021-03-16
Filed: 2020-01-08
Assignee: Disney Enterprises, Inc.
Inventors: Arana; Mark, Drake; Edward C., Farre Guiu; Miquel Angel, Accardo; Anthony M.
CPC: G06N3/08; G06N3/09; H04N21/23418; H04N21/251; H04N21/2541; H04N21/8352; H04N21/8358
Verdict: Set aside content authenticity forensics, business/anti-piracy
Source: Google Patents · FreePatentsOnline

Abstract

A system for assessing authenticity of modified content includes a computing platform having a hardware processor and a memory storing a software code including a neural network trained to assess the authenticity of modified content generated based on baseline digital content and including one or more modifications to the baseline digital content. The hardware processor executes the software code to use the neural network to receive the modified content and to assess the authenticity of each of the one or more modifications to the baseline digital content to produce one or more authenticity assessments corresponding respectively to the one or more modifications to the baseline digital content. The hardware processor is also configured to execute the software code to generate an authenticity evaluation of the modified content based on the one or more authenticity assessments, and to output the authenticity evaluation for rendering on a display.

Background

RELATED APPLICATION(S) (1) The present application is related to U.S. patent application Ser. No. 16/737,810, titled “Content Authentication Based on Intrinsic Attributes,” filed concurrently with the present application, and is hereby incorporated fully by reference into the present application. BACKGROUND (2) Advances in machine learning have enabled the production of realistic but forged recreations of a person's image or voice, known as “deepfakes” due to the use of deep artificial neural networks for their creation. Deepfakes may be produced without the consent of the person whose image or voice is being used, and may make the person being represented appear to say or do something that they have in fact not said or done. As a result, deepfake modified digital content can be used maliciously to spread misinformation. (3) Due to the widespread popularity of digital content for the distribution of entertainment and news, the effective authentication and management of that content is important to its creators, owners and distributors alike. However, as machine learning solutions continue to improve, deepfakes are and will continue to be difficult to detect. As a result, subtly modified or even entirely fake digital content may inadvertently be broadcast or otherwise distributed in violation of contractual agreement or regulatory restrictions, thereby subjecting the content owners and/or distributors to potential legal jeopardy. SUMMARY (4) There are provided systems and meth

Claims

1. A system comprising: a computing platform including a hardware processor and a system memory; a software code stored in the system memory, the software code including a neural network trained to assess an authenticity of a modified content generated based on a baseline digital content, the modified content including one or more modifications to the baseline digital content; the hardware processor being configured to execute the software code to: receive, using the neural network, the modified content; distinguish, using the neural network, the one or more modifications to the baseline digital content, from the baseline digital content; assess, using the neural network, an authenticity of each of the one or more modifications to the baseline digital content distinguished from the baseline digital content to produce one or more authenticity assessments corresponding respectively to the one or more modifications to the baseline digital content distinguished from the baseline digital content; generate an authenticity evaluation of the modified content based on the one or more authenticity assessments; and output the authenticity evaluation for rendering on a display. || 11. A method for use by a system including a computing platform having a hardware processor and a system memory storing a software code including a neural network trained to assess an authenticity of a modified content generated based on a baseline digital content, the modified content including one or more modifications to the baseline digital content, the method comprising: receiving, by the software code executed by the hardware processor and using the neural network, the modified content; distinguishing, by the software code executed by the hardware processor and using the neural network, the one or more modifications to the baseline digital content, from the baseline digital content; assessing, by the software code executed by the hardware processor and using the neural network, an authenticity of each of the one or more modifications to the baseline digital content distinguished from the baseline digital content to produce one or more authenticity assessments corresponding respectively to the one or more modifications to the baseline digital content distinguished from the baseline digital content; generating, by the software code executed by the hardware processor, an authenticity evaluation of the modified content based on the one or more authenticity assessments; and outputting, by the software code executed by the hardware processor, the authenticity evaluation for rendering on a display.