Abstract: A virtual accelerator is dynamically constructed on-demand to be used to execute one or more artificial intelligence tasks. The dynamically constructing the virtual accelerator on-demand includes determining, based on the one or more artificial intelligence tasks to be executed, computing resources to be allocated to the virtual accelerator on-demand. The computing resources are allocated to the virtual accelerator on-demand. The allocating of the computing resources includes assigning at least a portion of the computing resources to the virtual accelerator on-demand from standby computing resources. The virtual accelerator is to execute the one or more artificial intelligence tasks based on being allocated the computing resources.

Abstract: Secure verification of trained models, including: performing an inference operation on input inference data using a trained model; applying, to verifiable input data, one or more transformations based on an inference path of the trained model during the inference operation, thereby generating transformed verifiable input data; and verifying the trained model based on the transformed verifiable input data.

Abstract: Real-time validation of microcode updates may include: executing a workload on at least one first processor having a first version of a microcode instruction set; executing the workload on at least one second processor having a second version of a microcode instruction set, wherein the second version of the microcode instruction set comprises an update of the first version of the microcode instruction set; and validating the second version of the microcode instruction set based on a comparison of output from executing the workload using the at least one first processor and output from executing the workload using the at least one second processor.

Abstract: Method and apparatus for data backup. A first backup of a computing system is generated at a first time. A first confidence of compromise level of the computing system for the first time is generated. The first backup is stored along with metadata, where the metadata comprises the first confidence of compromise level of the computing system at the first time. In response to evaluating the first confidence of compromise level based on one or more backup criteria, a backup policy of the computing system is modified.

Abstract: Real-time validation of microcode updates may include: executing a workload on at least one first processor having a first version of a microcode instruction set; executing the workload on at least one second processor having a second version of a microcode instruction set, wherein the second version of the microcode instruction set comprises an update of the first version of the microcode instruction set; and validating the second version of the microcode instruction set based on a comparison of output from executing the workload using the at least one first processor and output from executing the workload using the at least one second processor.

Abstract: Method and apparatus for data backup. A first backup of a computing system is generated at a first time. A first confidence of compromise level of the computing system for the first time is generated. The first backup is stored along with metadata, where the metadata comprises the first confidence of compromise level of the computing system at the first time. In response to evaluating the first confidence of compromise level based on one or more backup criteria, a backup policy of the computing system is modified.