def get_ge(net_name, model_parameters, load_parameters):
args = util.EmptySpace()
for key, value in load_parameters.items():
setattr(args, key, value)
folder = "/media/rico/Data/TU/thesis/runs{}/{}".format(
args.experiment, util.generate_folder_name(args))
ge_x, ge_y = [], []
lta, lva, ltl, lvl = [], [], [], []
for run in runs:
filename = '{}/model_r{}_{}'.format(
folder, run, get_save_name(net_name, model_parameters))
ge_path = '{}.exp'.format(filename)
y_r = util.load_csv(ge_path, delimiter=' ', dtype=np.float)
x_r = range(len(y_r))
ge_x.append(x_r)
ge_y.append(y_r)
if show_losses or show_acc:
ta, va, tl, vl = util.load_loss_acc(filename)
lta.append(ta)
lva.append(va)
ltl.append(tl)
lvl.append(vl)
return ge_x, ge_y, (lta, lva, ltl, lvl)
def run_load(model, l2_penal, noise_level=-1.0, data_set=util.DataSet.RANDOM_DELAY_NORMALIZED):
args = util.EmptySpace()
args.use_hw = False
args.data_set = data_set
# args.traces_path = "/tudelft.net/staff-bulk/ewi/insy/CYS/spicek/student-datasets/"
# args.models_path = "/tudelft.net/staff-bulk/ewi/insy/CYS/spicek/rtubbing/"
args.traces_path = "/media/rico/Data/TU/thesis/data/"
args.models_path = "/media/rico/Data/TU/thesis/runs3/"
args.raw_traces = True
args.train_size = 40000
args.validation_size = 1000
args.attack_size = 9000
args.use_noise_data = True if noise_level > 0 else False
args.epochs = 75
args.batch_size = 100
args.lr = 0.0001
args.l2_penalty = float(l2_penal)
args.init_weights = "kaiming"
args.noise_level = float(noise_level)
args.type_network = 'HW' if args.use_hw else 'ID'
args.device = torch.device("cuda")
args.network_name = model
args.subkey_index = 2
args.unmask = True
args.desync = 0
args.spread_factor = 1
args.runs = 5
# args.kernels = [20]
# args.layers = [5]
args.kernels = [100, 50, 25, 20, 15, 17, 10, 7, 5, 3]
args.layers = [1, 2, 3, 4, 5, 6, 7, 8, 9]
args.channels = [32]
args.max_pool = 4
get_ranks(args)
x_exp, y_exp = test_with_key_guess_p(key_guesses_shuffled,
predictions_shuffled,
attack_size=args.attack_size,
real_key=real_key,
use_hw=args.use_hw)
y.append(y_exp)
# Calculate the mean over the experiments
y = np.mean(y, axis=0)
save_path = '{}model_r{}_{}_f{}.exp'.format(
folder, run, get_save_name(network_name, model_params), index_filter)
print("Save path {}".format(save_path))
save_np(save_path, y, f="%f")
load_args = util.EmptySpace()
load_args.device = util.device
load_args.models_path = "/media/rico/Data/TU/thesis/runs3"
load_args.use_hw = False
load_args.traces_path = "/media/rico/Data/TU/thesis/data/"
load_args.raw_traces = False
load_args.train_size = 40000
load_args.validation_size = 1000
load_args.attack_size = 5000
load_args.use_noise_data = False
load_args.data_set = util.DataSet.RANDOM_DELAY_NORMALIZED
load_args.subkey_index = 2
load_args.desync = 0
load_args.unmask = True
load_args.spread_factor = 1
load_args.epochs = 75
