Knox SDK frequently asked questions — ML model protection

For Linux, this depends on the environment settings. First, you need to check if execute permissions are present. If not, provide the permission using chmod 777 <tool> method.

No. The encrypted model runs the same way as a plain model, but with secure execution using Knox for Model Protection — Knox SDK APIs.

You can set the encrypted model saving location in the settings on the top right corner of the Encryption Tool.

You can set a location here if you want. Otherwise, the tool will save the encrypted file in a folder called ‘Encrypted Models’ in the same location where the plain ML file is located.

• Samsung Galaxy Xcover Pro (SM-G715), Galaxy Tab Active Pro (SM-T54x), Galaxy Xcover5(SM-G525), Galaxy Tab Active3 (SM-T57x) and Galaxy Xcover4s(SM-G398) with Knox SDK 3.7.1

• Galaxy S21/S21 FE, Galaxy Z Fold3, Galaxy S20/S20 FE, Galaxy Note20, Galaxy Z Fold 2, Galaxy Z Flip2, Galaxy Note 10*, Galaxy S10*, Galaxy Z Fold, Galaxy Z Flip/5G with Knox SDK 3.8

*Galaxy S10/Note 10 only support.KONPKG format

To see which tool version you can use depending on your device, refer to Knox ML Encryption Tool Revision page.

All major ML formats like Tensorflow (*.pb),Tensorflow Lite (*.tflite) and Caffe (*.caffemodel) are supported with Knox ML encryption tool V1.1 and from V1.2, coreML (*.mlmodel), keras (*.pb), pytorch (*.pth), ONNX (*.onnx) are also newly supported. To see the supported version of these formats, refer to Supported software packages in the Knox ML Encryption Tool page.

It is currently supported on Linux and Windows. The supported versions are:

• Windows — Windows 10
• Linux — Ubuntu 20.04