Hiding information and images via deep learning

A computing system comprising:

• an image hiding neural network configured to receive a package image and a cover image and, in response, output a carrier image that comprises the package image hidden within the cover image;
• one or more processors;
• and
• a first set of instructions that, when executed by the one or more processors, cause the computing system to:
• input the package image and the cover image to the image hiding neural network;
• and
• generate the carrier image that comprises the package image hidden within the cover image as an output of the image hiding neural network.

A computing system comprising:

• an image decoding neural network configured to receive a carrier image and, in response, output a reconstruction of a package image that was hidden within the carrier image;
• one or more processors;
• and
• a first set of instructions that, when executed by the one or more processors, cause the computing system to:
• input the carrier image to the image decoding neural network;
• and
• generate the reconstruction of the package image that comprises a representation of a hidden package image as an output of the image decoding neural network.

One or more non-transitory computer-readable media that store instructions that, when executed by one or more processors of a computing system, cause the computing system to:

• input a raw package image into an image preparation neural network;
• generate a prepared package image as an output of the image preparation neural network;
• input the prepared package image and a cover image into an image hiding neural network;
• generate a carrier image as an output of the image hiding neural network, the carrier image comprising the prepared package image hidden within the cover image;
• input the carrier image into an image decoding neural network;
• generate a reconstruction of the prepared package image as an output of the image decoding neural network;
• and
• simultaneously train the image decoding neural network based at least in part on a first loss function that describes a difference between the raw package image and the reconstruction of the raw package image, the image hiding neural network based at least in part on the first loss function and a second loss function that describes a difference between the cover image and the carrier image;
• and the image preparation neural network based at least in part on the first loss function and the second loss function.

A computer-implemented method to train neural networks for hiding information, the method comprising:

• inputting, by one or more computing devices, a package payload and a cover payload into an information hiding neural network;
• generating, by the one or more computing devices, a carrier payload as an output of the information hiding neural network, the carrier payload comprising the package payload hidden within the cover payload;
• inputting, by the one or more computing devices, the carrier payload into an information decoding neural network;
• generating, by the one or more computing devices, a reconstruction of the package payload as an output of the information decoding neural network;
• and
• simultaneously training, by the one or more computing devices, the information decoding neural network based at least in part on a first loss function that describes a difference between the package payload and the reconstruction of the package payload and the information hiding neural network based at least in part on the first loss function a second loss function that describes a difference between the cover payload and the carrier payload.