Encrypting AI Engine Kernels - 2023.2 English

AI Engine Tools and Flows User Guide (UG1076)

Document ID
Release Date
2023.2 English

The methodology for encrypting the AI Engine source code helps the IP developer to deliver IP sources to the end-user in a more secure way. The feature does not deliver AI Engine source file in clear text, and does not expose any kernel source content during compilation.

This feature aims for a trust model based upon the concerns of the IP author. The IP author determines how their secured IP can be viewed and used. The trust model can be summed up by the following: Encryption of source code is a mandate by the IP author that they expect their IP to be secure. Tool behavior defaults to the maximum reasonable protection that is available, while still accomplishing tool goals.

The AI Engine source encryption feature is available from a lounge website: https://www.xilinx.com/member/ai_engine_encryption.html (Registration required). Available on the lounge are the license and Synopsys keys that are necessary to encrypt and decrypt the AI Engine source code. Supporting documentation is also available.