Counter-ai camouflage
AssigneeRaytheon Company
InventorPhilip A. Sallee
Life of this patent
Estimating the term…
Abstract
Embodiments regard generating patterns that provide security from detection by artificial intelligence (AI) detection systems. A method includes generating, by a pattern generator, an image of camouflage, altering, by a camouflage painter, pixel values of a first target object in an image of the first target object to be consistent with the image of camouflage resulting in an image of a camouflaged target, concurrently determining (i) by a critic network, whether the image of the camouflaged target includes camouflage and (ii) by a target classifier whether the image of the camouflaged target includes the target object, and training the pattern generator to concurrently fool the critic network and the target classifier based on a combination of a loss of the critic network and a loss of the target classifier.
Claims20
A method for counter artificial intelligence (AI) camouflage, the method comprising:
- generating, by a pattern generator, an image of camouflage;
- altering, by a camouflage painter, pixel values of a first target object in an image of the first target object to be consistent with the image of camouflage resulting in an image of a camouflaged target;
- concurrently determining (i) by a critic network, whether the image of the camouflaged target includes camouflage and (ii) by a target classifier whether the image of the camouflaged target includes the target object;
- and
- training the pattern generator to concurrently fool the critic network and the target classifier based on a combination of a loss of the critic network and a loss of the target classifier.
A non-transitory machine-readable medium including instructions that, when executed by a machine, cause the machine to perform operations for counter artificial intelligence (AI) camouflage, the operations comprising:
- generating, by a pattern generator, an image of camouflage;
- altering, by a camouflage painter, pixel values of a target object in an image of the target object to be consistent with the image of camouflage resulting in an image of a camouflaged target;
- concurrently determining (i) by the critic network, whether the image of the camouflaged target includes camouflage and (ii) by the target classifier whether the image of the camouflaged target includes the target object;
- and
- training the pattern generator based on a combination of a loss of the critic network and a loss of the target classifier.
A system for counter artificial intelligence (AI) camouflage, the system comprising:
- processing circuitry;
- a memory including instructions that, when executed by the processing circuitry, cause the processing circuitry to perform operations comprising:
- generating, by a pattern generator, an image of camouflage;
- altering, by a camouflage painter, pixel values of a target object in an image of the target object to be consistent with the image of camouflage resulting in an image of a camouflaged target;
- concurrently determining (i) by the critic network, whether the image of the camouflaged target includes camouflage and (ii) by the target classifier whether the image of the camouflaged target includes the target object;
- and
- training the pattern generator based on a combination of a loss of the critic network and a loss of the target classifier.
Dependencies are inferred from the claim text.