diff options
Diffstat (limited to 'util/plot_gen.py')
| -rwxr-xr-x | util/plot_gen.py | 177 |
1 files changed, 113 insertions, 64 deletions
diff --git a/util/plot_gen.py b/util/plot_gen.py index 98d8261..0d518e6 100755 --- a/util/plot_gen.py +++ b/util/plot_gen.py @@ -12,21 +12,24 @@ # import numpy as np +from scipy.stats import entropy import matplotlib.pyplot as plt -import matplotlib.ticker as ticker -import matplotlib.colors as colors -from operator import itemgetter +from matplotlib import ticker, colors from copy import deepcopy import argparse +from utils import hw, moving_average, plot_hist + if __name__ == "__main__": parser = argparse.ArgumentParser(description="Plot results of ECTester key generation timing.") parser.add_argument("-o", "--output", dest="output", type=argparse.FileType("wb"), help="Write image to [file], do not display.", metavar="file") - parser.add_argument("--pub", dest="pub", action="store_true", help="Show public key scatter plot.") - parser.add_argument("--priv", dest="priv", action="store_true", help="Show private key scatter plot.") - parser.add_argument("--hist", dest="hist", action="store_true", help="Show histogram.") - parser.add_argument("--hw-hist", dest="hw_hist", action="store_true", help="Show Hamming weight 2D histogram (private key Hamming weight and generation time).") - parser.add_argument("--skip-first", dest="skip_first", action="store_true", help="Skip first entry, as it's usually a large outlier.") + parser.add_argument("--priv", dest="priv", action="store_true", help="Show private key MSB heatmap plot.") + parser.add_argument("--hist", dest="hist", action="store_true", help="Show keygen time histogram.") + parser.add_argument("--export-hist", dest="export_hist", action="store_true", help="Show export time histogram.") + parser.add_argument("--avg", dest="avg", action="store_true", help="Show moving average of keygen time.") + parser.add_argument("--hw-hist", dest="hw_hist", action="store_true", help="Show Hamming weight heatmap (private key Hamming weight and keygen time).") + parser.add_argument("--log", dest="log", action="store_true", help="Use logarithmic scale.") + parser.add_argument("--skip-first", dest="skip_first", nargs="?", const=1, type=int, help="Skip first entry, as it's usually a large outlier.") parser.add_argument("-t", "--title", dest="title", type=str, nargs="?", default="", help="What title to give the figure.") parser.add_argument("file", type=str, help="The file to plot(csv).") @@ -35,27 +38,53 @@ if __name__ == "__main__": with open(opts.file, "r") as f: header = f.readline() header_names = header.split(";") + if len(header_names) not in (4, 5): + print("Bad data?") + exit(1) - plots = [opts.priv, opts.pub, opts.hist, opts.hw_hist] + plots = [opts.priv, opts.hist, opts.export_hist, opts.avg, opts.hw_hist] n_plots = sum(plots) if n_plots == 0: - n_plots = 4 - plots = [True, True, True, True] + plots = [True for _ in range(5)] + if len(header_names) == 4: + n_plots = 4 + plots[2] = False + else: + n_plots = 5 + + + if plots[2] and len(header_names) != 5: + n_plots = n_plots - 1 + if n_plots == 0: + print("Nothing to plot.") + exit(1) + plots[2] = False hx = lambda x: int(x, 16) - data = np.genfromtxt(opts.file, delimiter=";", skip_header=1, converters={2: hx, 3: hx}, dtype=np.dtype([("index","u4"), ("time","u4"), ("pub", "O"), ("priv", "O")])) + if len(header_names) == 4: + data = np.genfromtxt(opts.file, delimiter=";", skip_header=1, converters={2: hx, 3: hx}, dtype=np.dtype([("index", "u4"), ("gen_time", "u4"), ("pub", "O"), ("priv", "O")])) + else: + data = np.genfromtxt(opts.file, delimiter=";", skip_header=1, converters={3: hx, 4: hx}, dtype=np.dtype([("index", "u4"), ("gen_time", "u4"), ("export_time", "u4"), ("pub", "O"), ("priv", "O")])) + if opts.skip_first: - data = data[1:] + data = data[opts.skip_first:] - if "nano" in header_names[1]: - unit = r"$\mu s$" - time_data = map(lambda x: x[1]//1000, data) - else: - unit = r"ms" - time_data = map(itemgetter(1), data) - time_data = list(time_data) - priv_data = list(map(itemgetter(2), data)) - pub_data = list(map(itemgetter(3), data)) + + gen_time_data = data["gen_time"] + export_time_data = None + if "export_time" in data.dtype.names: + export_time_data = data["export_time"] + pub_data = data["pub"] + priv_data = data["priv"] + + gen_unit = "ms" + if header_names[1].endswith("[nano]"): + gen_unit = r"$\mu s$" + np.floor_divide(gen_time_data, 1000, out=gen_time_data) + export_unit = "ms" + if len(header_names) == 5 and header_names[2].endswith("[nano]"): + export_unit = r"$\mu s$" + np.floor_divide(export_time_data, 1000, out=export_time_data) plt.style.use("ggplot") fig = plt.figure() @@ -66,66 +95,86 @@ if __name__ == "__main__": elif opts.title: fig.suptitle(opts.title) layout_kwargs["rect"] = [0, 0.02, 1, 0.98] - fig.tight_layout(**layout_kwargs) + fig.tight_layout(**layout_kwargs) + + max_gen_time = max(gen_time_data) + min_gen_time = min(gen_time_data) + bit_size = len(bin(max(priv_data))) - 2 + + sorted_data = np.sort(data, order="gen_time") + + i = 0 + entropies = {} + while i < len(data): + time_val = sorted_data["gen_time"][i] + j = i + msbs = [0 for _ in range(256)] + while j < len(data) and sorted_data["gen_time"][j] == time_val: + msbs[(sorted_data["priv"][j] >> (bit_size - 8)) & 0xff] += 1 + j += 1 + if j - 100 > i: + entropies[time_val] = entropy(msbs, base=2) + i = j + + entropy = sum(entropies.values())/len(entropies) + + cmap = deepcopy(plt.cm.plasma) + cmap.set_bad("black") + + norm = colors.Normalize() + if opts.log: + norm = colors.LogNorm() plot_i = 1 if plots[0]: axe_private = fig.add_subplot(n_plots, 1, plot_i) - axe_private.scatter(time_data, priv_data, marker="x", s=10) - axe_private.set_ylabel("private key value\n(big endian)") - axe_private.set_xlabel("time ({})".format(unit)) + priv_msb = np.array(list(map(lambda x: x >> (bit_size - 8), priv_data)), dtype=np.dtype("u1")) + max_msb = max(priv_msb) + min_msb = min(priv_msb) + heatmap, xedges, yedges = np.histogram2d(priv_msb, gen_time_data, bins=[max_msb - min_msb, max_gen_time - min_gen_time]) + extent = [min_msb, max_msb, yedges[0], yedges[-1]] + axe_private.imshow(heatmap.T, extent=extent, aspect="auto", cmap=cmap, origin="low", interpolation="nearest", norm=norm) + axe_private.set_xlabel("private key MSB value") + axe_private.set_ylabel("keygen time ({})".format(gen_unit)) plot_i += 1 if plots[1]: - axe_public = fig.add_subplot(n_plots, 1, plot_i) - axe_public.scatter(time_data, pub_data, marker="x", s=10) - axe_public.set_ylabel("public key value\n(big endian)") - axe_public.set_xlabel("time ({})".format(unit)) + axe_hist = fig.add_subplot(n_plots, 1, plot_i) + plot_hist(axe_hist, gen_time_data, "keygen time ({})".format(gen_unit), opts.log) plot_i += 1 if plots[2]: axe_hist = fig.add_subplot(n_plots, 1, plot_i) - time_max = max(time_data) - time_avg = np.average(time_data) - time_median = np.median(time_data) - axe_hist.hist(time_data, bins=time_max//3, log=True) - axe_hist.axvline(x=time_avg, alpha=0.7, linestyle="dotted", color="red", label="avg = {}".format(time_avg)) - axe_hist.axvline(x=time_median, alpha=0.7, linestyle="dotted", color="green", label="median = {}".format(time_median)) - axe_hist.set_ylabel("count\n(log)") - axe_hist.set_xlabel("time ({})".format(unit)) - axe_hist.xaxis.set_major_locator(ticker.MaxNLocator()) - axe_hist.legend(loc="best") + plot_hist(axe_hist, export_time_data, "export time ({})".format(export_unit), opts.log) plot_i += 1 if plots[3]: - priv_bit_bins = {} - for i in range(len(data)): - skey = priv_data[i] - time = time_data[i] - skey_hw = 0 - while skey: - skey_hw += 1 - skey &= skey - 1 - if skey_hw in priv_bit_bins: - priv_bit_bins[skey_hw].append(time) - else: - priv_bit_bins[skey_hw] = [time] - priv_bit_x = [] - priv_bit_y = [] - for k,v in priv_bit_bins.items(): - priv_bit_x.extend([k] * len(v)) - priv_bit_y.extend(v) + axe_avg = fig.add_subplot(n_plots, 1, plot_i) + #if len(header_names) == 5: + # axe_other = axe_avg.twinx() + # axe_other.plot(moving_average(export_time_data, 100), color="green", alpha=0.6, label="export, window = 100") + # axe_other.plot(moving_average(export_time_data, 1000), color="yellow", alpha=0.6, label="export, window = 1000") + # axe_other.legend(loc="lower right") + axe_avg.plot(moving_average(gen_time_data, 100), label="window = 100") + axe_avg.plot(moving_average(gen_time_data, 1000), label="window = 1000") + axe_avg.set_ylabel("keygen time ({})".format(gen_unit)) + axe_avg.set_xlabel("index") + axe_avg.legend(loc="best") + plot_i += 1 + + if plots[4]: axe_priv_hist = fig.add_subplot(n_plots, 1, plot_i) - h, xe, ye = np.histogram2d(priv_bit_x, priv_bit_y, bins=[max(priv_bit_bins) - min(priv_bit_bins), (max(time_data) - min(time_data))//5]) - cmap = deepcopy(plt.cm.plasma) - cmap.set_bad("black") - im = axe_priv_hist.imshow(h.T, origin="low", cmap=cmap, aspect="auto", extent=[xe[0], xe[-1], ye[0], ye[-1]], norm=colors.LogNorm()) + priv_hw = np.array(list(map(hw, priv_data)), dtype=np.dtype("u2")) + h, xe, ye = np.histogram2d(priv_hw, gen_time_data, bins=[max(priv_hw) - min(priv_hw), max_gen_time - min_gen_time]) + im = axe_priv_hist.imshow(h.T, origin="low", cmap=cmap, aspect="auto", extent=[xe[0], xe[-1], ye[0], ye[-1]], norm=norm) + axe_priv_hist.axvline(x=bit_size//2, alpha=0.7, linestyle="dotted", color="white", label=str(bit_size//2) + " bits") axe_priv_hist.set_xlabel("private key Hamming weight") - axe_priv_hist.set_ylabel("time ({})".format(unit)) + axe_priv_hist.set_ylabel("keygen time ({})".format(gen_unit)) + axe_priv_hist.legend(loc="best") fig.colorbar(im, ax=axe_priv_hist) - if plot_i > 2: - fig.text(0.01, 0.02, "Data size: {}".format(len(time_data)), size="small") + fig.text(0.01, 0.02, "Data size: {}".format(len(gen_time_data)), size="small") + fig.text(0.01, 0.04, "Entropy of privkey MSB(estimated): {:.2f} b".format(entropy), size="small") if opts.output is None: plt.show() |