HDF5 container¶

When performing side-channel analysis, a user will seek to store its data for further usages. To our knowledge, every lab has its own format for storing side-channel traces.

The probably most popular traces storage file format is HDF5 (Hierarchical Data Format). For each set of traces, a single HDF5 file is created, implementing two datasets: the first one for leakages, and the second for values.

Lascar provides the class Hdf5Container which allows handling this format transparently. This enables working on traces without having to load them all in memory (as a TraceBatchContainer would do).

The Hdf5Container.export method can be used to convert any container into a Hdf5Container (i.e. by creating a file containing all its traces). In the following example, a TraceBatchContainer is created, and converted into a Hdf5Container.

from lascar import TraceBatchContainer, Hdf5Container
import numpy as np

leakages = np.random.rand(10, 100)  # 10 leakages, 100 time samples each.
values = np.random.randint(0, 256, (10, 16))  # 10 associated values, 16 bytes each.
batch = TraceBatchContainer(leakages, values)
hdf5 = Hdf5Container.export(batch, "tmp.h5")

Since Hdf5Container is a Container, it implements all the features seen before. This can be verified as follows:

for (a, b) in zip(batch, hdf5, strict=True)
    assert a == b