Methods and systems for protecting digital content against artificial intelligence-based unauthorized manipulation of the digital content
AssigneeAccenture Global Solutions Limited
InventorsLeah Ding, Xiaoyong Yuan, Malek Ben Salem
Life of this patent
Estimating the term…
Abstract
A device may receive digital content and may process the digital content, with at least one of an optimization-based poisoning model or a statistical-based poisoning model, to generate at least one of first poisoning data or second poisoning data, respectively. The device may generate new digital content based on the digital content and the at least one of the first poisoning data or the second poisoning data. The device may provide the new digital content to one or more devices to be accessed by at least one deepfake model used to create fake digital content and may perform one or more actions based on the new digital content.
Claims19
A method, comprising:
- receiving, by a device, digital content;
- processing, by the device, the digital content, with at least one of an optimization-based poisoning model or a statistical-based poisoning model, to generate at least one of first poisoning data or second poisoning data, respectively;
- generating, by the device, new digital content based on the digital content and the at least one of the first poisoning data or the second poisoning data;
- providing, by the device, the new digital content to one or more devices to be accessed by at least one deepfake model used to create fake digital content;
- and
- performing, by the device, one or more actions based on the new digital content,
- wherein performing the one or more actions based on the new digital content comprises one or more of:
- causing generation of degraded fake digital content by the at least one deepfake model based on the new digital content;
- or
- preventing the at least one deepfake model from generating the fake digital content based on the new digital content.
A device, comprising:
- one or more memories;
- and
- one or more processors, communicatively coupled to the one or more memories, configured to:
- receive digital content;
- process the digital content, with at least one of an optimization-based poisoning model or a statistical-based poisoning model, to generate at least one of first poisoning data or second poisoning data, respectively;
- generate new digital content based on the digital content and the least one of the first poisoning data or the second poisoning data;
- provide the new digital content to one or more devices to be accessed by at least one artificial intelligence model used to create fake digital content;
- and
- perform one or more actions based on the new digital content,
- wherein the one or more processors, when performing the one or more actions based on the new digital content, are configured to one or more of:
- prevent the at least one artificial intelligence model from removing the first poisoning data or the second poisoning data from the new digital content;
- or
- retrain the optimization-based poisoning model or the statistical-based poisoning model based on the new digital content.
A non-transitory computer-readable medium storing a set of instructions, the set of instructions comprising:
- one or more instructions that, when executed by one or more processors of a device, cause the device to:
- receive digital content;
- process the digital content, with at least one of an optimization-based poisoning model or a statistical-based poisoning model, to generate at least one of first poisoning data or second poisoning data, wherein the process the digital content comprises:
- receiving a sample taken from a multivariate normal distribution with a mean and a covariance estimated from the digital content;
- rounding feature values of the sample to identify a portion of the sample that poisons pixels of the digital content;
- selectin a response variable value of the sample to identify pixels of the digital content to poison;
- and
- poisoning the identified pixels of the digital content with the portion of the sample to generate the second poisoning data;
- generate new digital content based on the digital content and the at least one of the first poisoning data or the second poisoning data;
- provide the new digital content to one or more devices to be accessed by at least one deepfake model, used to create fake digital content;
- and
- cause generation of degraded fake digital content by the at least one deepfake model.
Dependencies are inferred from the claim text.