In one embodiment, a computer-implemented method of a data processing (DP) accelerator obtaining a watermark of a watermark-enable artificial intelligence (AI) model includes receiving, by the DP accelerator, input data to the DP accelerator that causes the watermark-enabled AI model to extract the watermark from the watermark-enabled AI model; and providing the watermark of the watermark-enabled AI model to the host device. The DP accelerator can receive the model from the host device. The DP accelerator can further receive a command to digitally sign the watermark and call a security unit of the DP accelerator to digitally sign the watermark.

In one embodiment, a computer-implemented method of a data processing (DP) accelerator obtaining a watermark of a watermark-enable artificial intelligence (AI) model includes receiving, by the DP accelerator, input data to the DP accelerator that causes the watermark-enabled AI model to extract the watermark from the watermark-enabled AI model; and providing the watermark of the watermark-enabled AI model to the host device. The DP accelerator can receive the model from the host device. The DP accelerator can further receive a command to digitally sign the watermark and call a security unit of the DP accelerator to digitally sign the watermark.

Embodiments are described for placing a watermark over application windows in a desktop. For each application window that is opened in the desktop, the system can determine whether the application requires a watermark, for example, based on a predefined list that specifies which applications require watermarks. For each application window that requires a watermark, a uncovered watermark region can be calculated where the watermark will appear. An overlay can be placed over the application windows, for example in a top-level window that does not receive mouse and keyboard inputs, and the watermark can be drawn in the overlay over the location of the uncovered watermark region of each application. As a result, watermarks can be placed over a plurality of specified application windows in an efficient and convenient manner.

Disclosed herein is a system for watermarking a neural network, comprising memory; and at least one processor in communication with the memory; wherein the memory stores instructions for causing the at least one processor to carry out a method comprising: generating a trigger set by obtaining examples from a training set by random sampling from the training set, respective examples being associated with respective true classes of a plurality of classes; generating a set of adversarial examples by structured perturbation of the examples; generating, for each adversarial example, one or more adversarial class labels by passing the adversarial example to the neural network; and applying one or more trigger labels to each said adversarial example, wherein the one or more trigger labels are selected randomly from the plurality of classes, and wherein each trigger label is not a said true class label for the corresponding example or a said adversarial class label for the corresponding adversarial example; and storing the adversarial examples and corresponding trigger labels as the trigger set; and performing a tuning process to adjust parameters at each layer of the neural network using the trigger set, to thereby generate a watermarked neural network.

Disclosed herein is a system for watermarking a neural network, comprising memory; and at least one processor in communication with the memory; wherein the memory stores instructions for causing the at least one processor to carry out a method comprising: generating a trigger set by obtaining examples from a training set by random sampling from the training set, respective examples being associated with respective true classes of a plurality of classes; generating a set of adversarial examples by structured perturbation of the examples; generating, for each adversarial example, one or more adversarial class labels by passing the adversarial example to the neural network; and applying one or more trigger labels to each said adversarial example, wherein the one or more trigger labels are selected randomly from the plurality of classes, and wherein each trigger label is not a said true class label for the corresponding example or a said adversarial class label for the corresponding adversarial example; and storing the adversarial examples and corresponding trigger labels as the trigger set; and performing a tuning process to adjust parameters at each layer of the neural network using the trigger set, to thereby generate a watermarked neural network.

A software and/or hardware facility that can be used by content owners to assert ownership of content so that copyright friendly websites and services can take action against copyright piracy effectively, efficiently and is scalable is disclosed. The facility makes available to all content owners watermarking/fingerprinting technology so an identifier (e.g., a unique code) can be embedded in the content (e.g., video/audio portion of each video content asset). The facility utilizes blockchain technology to add information related to each unique identifier in a database and allows an authorized user (e.g., the owner) to update the information through a blockchain transaction.

A software and/or hardware facility that can be used by content owners to assert ownership of content so that copyright friendly websites and services can take action against copyright piracy effectively, efficiently and is scalable is disclosed. The facility makes available to all content owners watermarking/fingerprinting technology so an identifier (e.g., a unique code) can be embedded in the content (e.g., video/audio portion of each video content asset). The facility utilizes blockchain technology to add information related to each unique identifier in a database and allows an authorized user (e.g., the owner) to update the information through a blockchain transaction.

Using various embodiments, methods and systems for verification of a digital asset owner in a digital environment are described. In one embodiment, a system is configured to receive a non-fungible token (NFT) associated with a digital asset, the NFT providing proof of ownership of the digital asset through a cryptographic public key and retrieve the digital asset. The system then retrieves a secret pattern from the digital asset, wherein the secret pattern was previously embedded into the digital asset, the secret pattern associated with the cryptographic public key and computes a first identification hash value using a hash function, the hash function receiving a parameter value derived from the secret pattern. The system then receives a second identification hash value and compares the first identification hash value to the second identification hash value. If the first and second identification hash values are identical, then the digital asset is determined to be authentic.

Using various embodiments, methods and systems for verification of a digital asset owner in a digital environment are described. In one embodiment, a system is configured to receive a non-fungible token (NFT) associated with a digital asset, the NFT providing proof of ownership of the digital asset through a cryptographic public key and retrieve the digital asset. The system then retrieves a secret pattern from the digital asset, wherein the secret pattern was previously embedded into the digital asset, the secret pattern associated with the cryptographic public key and computes a first identification hash value using a hash function, the hash function receiving a parameter value derived from the secret pattern. The system then receives a second identification hash value and compares the first identification hash value to the second identification hash value. If the first and second identification hash values are identical, then the digital asset is determined to be authentic.

A computer-implemented system for associating digital media files with external commodities wherein a digital media file such as digital art is provided. An external commodity that is unique and has a unique identifier is provided, such as a carbon credit. Ownership of the external commodity is transferred to a buyer when the digital media file is acquired by the buyer. The digital media file is associated with the external commodity by generating a digital signature using a private key from data that includes a hash of the digital media file and the unique commodity identifier. A new, unique digital file is created containing the digital signature as metadata.