[analysis] Fix num_traces in TVLA plot

[nasahlpa]

Previously, when number_of_steps > 1, the TVLA plot only contained the number of traces used for a step, but not the total number of traces processed. This PR fixes the issue, now the total number of used traces is printed in the figure.

[vrozic]

I'm not sure that this value should always be zero.
The condition of this if branch is:

if (cfg["input_histogram_file"] is None or cfg["output_histogram_file"] is not None) \
            and cfg["ttest_step_file"] is None:

so it is possible that cfg["input_histogram_file"] is not None (e.g. when we specify both input and output histogram files),
so the initial value for num_traces_used_total should be the number of traces used in the input histogram file.

[nasahlpa]

Good point, thanks! However, do we actually read the number of traces of the histogram when cfg["input_histogram_file"] is provided? Should this be done here:

ot-sca/analysis/tvla.py

Line 328 in 6bbb5ac

if cfg["input_histogram_file"] is not None:

[vogelpi]

I think you're both somewhat right here. The histograms are defined as follows:

                # For every time sample we make two histograms, one for Hamming weight of the
                # sensitive variable = 0 (fixed set) and one for Hamming weight > 0 (random set).
                # histograms has dimensions [num_rnds, num_bytes, 2, num_samples, trace_resolution]
                # The value stored in histograms[v][w][x][y][z] shows how many traces have value z
                # at sample y, given that HW(state byte w in AES round v) = 0 (fixed set, x = 0) or
                # > 0 (random set, x = 1).

When an input histograms file is used, the number of traces used can be computed as sum(histograms[:][:][:][:][0][:]).

Since we already didn't set num_traces_used_total correctly when loading histograms before this PR. I suggest to create a new issue for tracking and merge this on as is.

[vrozic]

I agree that this can be merged as is.
I've added an item to issue #72 to track this.

[nasahlpa]

Thanks to both of you!

[vogelpi]

LGTM, I suggest to create an issue for tracking the handling of the histogram loading feature.