lascar-AES-HOCPA.py

from lascar import *
from lascar.tools.aes import sbox

filename = "/home/johndoe/ASCAD_data/ASCAD_databases/ATMega8515_raw_traces.h5"

# poi = Points Of Interest.
# We use the SNRs computed before to extract time samples where both mask and masked sensitive variable are used.
snrs = DictOutputMethod.load('SNR.pickle') # we load the snr results computed at 02-snr.py
poi = []
poi = [snrs['SNR4: masked sbox output in linear parts'][5000].argmax()]
poi += [snrs['SNR5: sbox output mask in linear parts'][5000].argmax()]

# Container creation: we take the same file, but set the leakage_section to the computed poi.
container = Hdf5Container(filename,
                          leakages_dataset_name="traces",
                          values_dataset_name="metadata",
                          leakage_section=poi,
                          )

container.number_of_traces = 500

# Then we make a container that will apply a CenteredProduct to recombine all the points of interest.
container.leakage_processing = CenteredProductProcessing(container)

# Now a classical CPA, targetting the output of the 3rd Sbox:
cpa_engine = CpaEngine("cpa-high-order",
                       lambda value, guess: hamming(sbox[value['plaintext'][3] ^ guess]),
                       range(256),
                       solution=container[0].value['key'][3],
                       jit=False)

session = Session(container,
                  engine=cpa_engine,
                  output_method=ScoreProgressionOutputMethod(cpa_engine),
                  output_steps=10) # the steps at which the results will be computed

session.run(1000)