This application is provided as an example of how to use TrustZone to separate the inference from an application layer in secure and nonsecure world. This software is primarily intended for guidance, to demonstrate how the firmware can be partitioned.
Building with default settings requires CMake for the configuration and make for building. This will produce an elf file which can be run on the FVP.
$ cmake -B build targets/corstone-300 $ make -C build trustzone
The model in this example has been generated from
image_classification/mobilenet_v2_1.0_224/tflite_uint8
Found in the ARM ML-zoo open source github repository:
https://github.com/ARM-software/ML-zoo.git
The model is converted using vela (v2.0.1)
vela --cascading=True --config vela.ini --system-config=Ethos_U55_High_End_Embedded --memory-mode=Shared_Sram --accelerator-config=ethos-u55-128 --block-config-limit=0 ML-zoo/models/image_classification/mobilenet_v2_1.0_224/tflite_uint8/mobilenet_v2_1.0_224_quantized_1_default_1.tflite
Where the vela.ini is fetched from the vela git repository:
wget https://review.mlplatform.org/plugins/gitiles/ml/ethos-u/ethos-u-vela/+/refs/heads/master/vela.ini
Then the compiled model representation has been converted to a header file using xxd.
xxd -i output/mobilenet_v2_1.0_224_quantized_1_default_1_vela.tflite > model.h
The input and expected output data has been taken from the ML-zoo and generated from the respective .npy files to header files with:
python3 -c 'import numpy;i=numpy.load("ML-zoo/models/image_classification/mobilenet_v2_1.0_224/tflite_uint8/testing_input/input/0.npy");i.tofile("input.bin");' && xxd -i input.bin > input.h
python3 -c 'import numpy;i=numpy.load("ML-zoo/models/image_classification/mobilenet_v2_1.0_224/tflite_uint8/testing_output/output/0.npy");i.tofile("output.bin");' && xxd -i output.bin > output.h
The generated header files are then modified to adapt name and link attributes.
Assuming that the Corstone-300 FVP has been downloaded, installed and placed in the PATH variable. Then the software binary can be tested like this.
$ ctest trustzone
Note! The expected output using the input data is not bitexact.
The Arm Ethos-U Core Platform is provided under an Apache-2.0 license. Please see LICENSE.txt for more information.
Please see README
Please see Security.