System Protection Application for Foundation Model Appliances
In AI foundation model inference scenarios, stricter regulations and increasingly sophisticated attacks are driving higher demands for security in both the execution environment and the models themselves.
The Kunpeng BMC chip provides a hardware RoT that verifies each layer of firmware through signature verification, protecting the system and firmware from being tampered with. Runtime code is continuously measured to monitor and control the execution environment, preventing malicious injection or modification during operation. A trustlist of hardware and software components blocks unauthorized access or changes, further safeguarding model integrity. In addition, customer-supplied keys are combined with the hardware RoT to encrypt models and private static data, protecting valuable assets and sensitive information from leakage or misuse.
