Repository files navigation Artifact for 'Activation Functions Considered Harmful: Recovering Neural Network Weights through Controlled Channels' Instructions for MNIST example
Setup sgx-step as indicated in README-sgxstep.md
You will need SGX-capable hardware
build tensorflowlite-micro
cd app/nn_param_recovery/Enclave/tflite-micromake -f ./tensorflow/lite/micro/tools/make/Makefile BUILD_TYPE=releaseYou should build './gen/linux_x86_64_release_gcc/lib/libtensorflow-microlite.a'
Train the victim model:
cd app/nn_param_recovery/models/mnistpython3 train_mnist.pypython3 test_mnist.py
Convert the tf-lite model into its header file representation:
cd ../python3 generate_cc_arrays.py . mnist/models/mnist.tflite
Start the attack (this will build the victim enclave):
cd app/nn_param_recovery/python3 attack.py
The attack will create a .pickle file to checkpoint itself as it works.
When all of the convergence points have been found, a table will be printed which should match the table in the paper.
You can’t perform that action at this time.
