EMMA is a framework for capturing and attacking traces of electromagnetic radiation emitted by an electronic device, in order to obtain encryption keys or other sensitive data.

The following Python 3 packages are required by EMMA: celery, tensorflow, keras, matplotlib, numpy, sklearn.

Two config files should be added to the EMMA root directory: settings.conf and datasets.conf.

[Network]
broker = redis://:password@redisserver:6379/0
backend = redis://:password@redisserver:6379/0

[Datasets]
datasets_path = /home/user/my-dataset-directory/
stream_interface = eth0

[EMMA]
remote = True

# Custom dataset (train)
[em-corr-arduino]
format=cw
reference_index=0

# Custom dataset (test)
[em-cpa-arduino]
format=cw
reference_index=0

# ASCAD database
[ASCAD]
format=ascad
reference_index=0

Although EMMA can run on a single machine, ideally at least two machines should be used to perform an analysis: a master (low-end device for sending commands) and one or more slaves (high-end devices for performing calculations). At least one device must have redis installed and configured. Before continuing, make sure this device is listed as the backend in settings.conf (see above).

On each slave that should perform computations, run the following command to automatically spawn worker processes for each available CPU:

celery -A emma_worker worker -l info -Q celery,priority.high

Note that each slave should be able to access the datasets listen in datasets.conf. Now, the master can issue commands to process these datasets.

emma.py plot ASCAD

See emma.py -h for a full list of available commands.

EMcap is a tool that allows for convenient capturing and storage of EM trace datasets using Software Defined Radios (SDRs).