利用圆形棒棒糖图探索密码强度与流行度

HsuHeinrich

发布于 2025-10-20 18:27:56

1140

利用圆形棒棒糖图探索密码强度与流行度

import matplotlib.pyplot as plt
import numpy as np
import pandas as pd

数据探索

以下数据如果有需要的同学可关注公众号HsuHeinrich，回复【数据可视化】自动获取～

df_pw = pd.read_csv('https://raw.githubusercontent.com/rfordatascience/tidytuesday/master/data/2020/2020-01-14/passwords.csv')
df_pw = df_pw.dropna(subset=['password'])
df_pw.head()

image-20240129180657919

rank：在已发布密码库中的排名 password：密码的文本 category：密码所属的类别 value：通过在线猜测破解的时间 time_unit：与value匹配的时间单位

def to_seconds(value, time_unit):
    '''
    将时间转换为统一的秒
    '''
    if time_unit == "seconds":
        return value
    elif time_unit == "minutes":
        return value * 60
    elif time_unit == "hours":
        return value * 60 * 60
    elif time_unit == "days":
        return value * 60 * 27
    elif time_unit == "weeks":
        return value * 60 * 24 * 7
    elif time_unit == "months":
        return value * 60 * 24 * 30
    elif time_unit == "years":
        return value * 60 * 24 * 365
    else:
        return np.nan

# 时间列表（秒）
TIMES = [
    to_seconds(row["value"], row["time_unit"])
    for _, row in df_pw.iterrows()
]

TIME_MAX = np.max(TIMES)
TIME_MIN = np.min(TIMES)

# 时间缩放函数
def scale_to_interval(x, low=1, high=60):
    return ((x - TIME_MIN) / (TIME_MAX - TIME_MIN)) * (high - low) + low

绘制基本圆形图

# 基础变量设置

# 定义多种不同的灰色
GREY88 = "#e0e0e0"
GREY85 = "#d9d9d9"
GREY82 = "#d1d1d1"
GREY79 = "#c9c9c9"
GREY97 = "#f7f7f7"
GREY60 = "#999999"

# x轴的值
ANGLES = np.linspace(0, 2 * np.pi, len(TIMES), endpoint=False)

# y的长度
HEIGHTS = np.array(TIMES)

# 类别编码
CATEGORY_CODES = pd.Categorical(df_pw["category"]).codes

# 自定义Colormap
COLORMAP = ["#5F4690", "#1D6996", "#38A6A5", "#0F8554", "#73AF48", 
            "#EDAD08", "#E17C05", "#CC503E", "#94346E", "#666666"]

# 为每个密码类别匹配颜色
COLORS = np.array(COLORMAP)[CATEGORY_CODES]


# 常量，用于提升空间距离（轴心到起始点）
PLUS = 1000

# 存储指定的密码信息：value>90（结果包含4个密码）
LABELS_DF = df_pw[df_pw["value"] > 90].reset_index()
# 增加标签列
LABELS_DF["label"] = [
    f"{pswrd}\nRank: {int(rank)}" 
    for pswrd, rank in zip(LABELS_DF["password"], LABELS_DF["rank"])
]

# 设置指定密码的位置信息

# x的位置为TIMES对应的索引，偏差是为了插入文本标签
LABELS_DF["x"] = [40, 332, 401, 496] 
# y的长度为HEIGHTS的长度，因为y轴对数化了，所以48498112和160000000转化为对数基本无差
LABELS_DF["y"] = [160000000, 90000000, 45000000, 48498112]

# 初始化布局
fig, ax = plt.subplots(figsize=(8, 8), subplot_kw={"projection": "polar"})

# 背景色
fig.patch.set_facecolor("white")
ax.set_facecolor("white")

# 径向轴刻度对数化
ax.set_rscale('symlog')

# 角度起始位置：90度
ax.set_theta_offset(np.pi / 2)

# 反转方向：按逆时针旋转绘制
ax.set_theta_direction(-1)

# 添加线
ax.vlines(ANGLES, 0 + PLUS, HEIGHTS + PLUS, color=COLORS, lw=0.9)

# 添加点
ax.scatter(ANGLES, HEIGHTS + PLUS, s=scale_to_interval(HEIGHTS), color=COLORS);

output_12_0

删除不必要的元素

# 移除边框（极坐标）
ax.spines["start"].set_color("none")
ax.spines["polar"].set_color("none")

# 移除网格线、x刻度、y刻度标签
ax.grid(False)
ax.set_xticks([])
ax.set_yticklabels([])

# 添加自定义网格线：每圈分别代表一天、周、月、年
HANGLES = np.linspace(0, 2 * np.pi, 200)
ax.plot(HANGLES, np.repeat(1 * 24 * 60 + PLUS, 200), color= GREY88, lw=0.7)
ax.plot(HANGLES, np.repeat(7 * 24 * 60 + PLUS, 200), color= GREY85, lw=0.7)
ax.plot(HANGLES, np.repeat(30 * 24 * 60 + PLUS, 200), color= GREY82, lw=0.7)
ax.plot(HANGLES, np.repeat(365 * 24 * 60 + PLUS, 200), color= GREY79, lw=0.7)

# 为指定的密码添加标签
for idx, row in LABELS_DF.iterrows():
    color = COLORS[row["index"]]
    ax.text(
        x=ANGLES[row["x"]], y=row["y"], s=row["label"], color=color,
        ha="right", va="center", ma="center", size=8,
        family="Roboto Mono", weight="bold"
    )

fig

output_14_0

丰富信息

# 灰色填充
ax.fill(HANGLES, np.repeat(PLUS, 200), GREY97)

# 添加文本信息
# 利用transform=ax.transAxes根据极坐标位置添加标签。例如(0.5, 0.5)表示中间位置
ax.text(
    x=0.5, y=0.58, s="********\nCracking\nYour Favorite\nPassword\n********",
    color=GREY60, va="center", ha="center", ma="center", fontfamily="Roboto Mono",
    fontsize=18, fontweight="bold", linespacing=0.87, transform=ax.transAxes
)

ax.text(
    x=0.5, y=0.46, s="Time it takes to crack the 500 most\ncommon passwords by online guessing.\nSorted by rank and colored by category.",
    color=GREY60, va="center", ha="center",  ma="center", fontfamily="Roboto Mono",
    fontsize=7, linespacing=0.87, transform=ax.transAxes
)

ax.text(
    x=0.5, y=0.39, s="Time is displayed on a logarithmic scale\nwith the rings representing one day,\none week, one month, and one year\n(from inner to outer ring).",
    color=GREY60, va="center", ha="center",  ma="center", fontfamily="Roboto Mono",
    fontsize=7, linespacing=0.87, transform=ax.transAxes
)

fig

绘制多个子图

为每个类别的密码单独绘制圆形棒棒糖图

# 将上述的步骤抽象为函数circular_plot
def circular_plot(angles, heights, colors, lw, ax):
    ax.set_facecolor("white")
    
    ax.set_rscale("symlog")
    ax.set_theta_offset(np.pi / 2)
    ax.set_theta_direction(-1)
    
    ax.spines["start"].set_color("none")
    ax.spines["polar"].set_color("none")
    
    ax.grid(False)
    ax.set_xticks([])
    ax.set_yticklabels([])
    
    # lw用来控制线宽
    ax.vlines(angles, 0 + PLUS, heights + PLUS, color=colors, lw=lw)
    ax.scatter(angles, heights + PLUS, s=scale_to_interval(heights), color=colors)
    
    HANGLES = np.linspace(0, 2 * np.pi, 200)
    ax.plot(HANGLES, np.repeat(1 * 24 * 60 + PLUS, 200), color= GREY88, lw=0.7)
    ax.plot(HANGLES, np.repeat(7 * 24 * 60 + PLUS, 200), color= GREY85, lw=0.7)
    ax.plot(HANGLES, np.repeat(30 * 24 * 60 + PLUS, 200), color= GREY82, lw=0.7)
    ax.plot(HANGLES, np.repeat(365 * 24 * 60 + PLUS, 200), color= GREY79, lw=0.7)
    
    ax.fill(HANGLES, np.repeat(PLUS, 200), GREY97)
    
    # 扩大极坐标半径：给予较大点更为适合的绘图区域
    ax.set_rmax(ax.get_rmax() * 2)

# 自定义函数map_category，将一些名称较长的category换为换行符
def map_category(category):
    if category == "cool-macho":
        return "cool-\nmacho"
    elif category == "nerdy-pop":
        return "nerdy-\npop"
    elif category == "password-related":
        return "password-\nrelated"
    elif category == "rebellious-rude":
        return "rebel-\nlious-\nrude"
    elif category == "simple-alphanumeric":
        return "simple-\nalpha-\nnumeric"
    else:
        return category

# 类别标签
CATEGORIES = sorted(pd.Categorical(df_pw["category"]).unique())
LABELS = [map_category(category) for category in CATEGORIES]

# 绘制多子图：2*5
fig, axes = plt.subplots(2, 5, figsize=(15, 6), subplot_kw={"projection": "polar"})
fig.patch.set_facecolor("white")

# 定义子图的位置
SLICES = [(i, j) for i in range(2) for j in range(5)]

for category, label, slice_ in zip(CATEGORIES, LABELS, SLICES):
    # 子图位置
    ax = axes[slice_]
    
    # 指定类别的索引位置（列表形式）
    idx = df_pw.index[df_pw["category"] == category].tolist()
    
    # 筛选该类比的所有数据：ANGLES、HEIGHTS、COLORS
    angles = ANGLES[idx]
    heights = HEIGHTS[idx]
    colors = COLORS[idx]
    
    # 绘图
    circular_plot(angles, heights, colors, 0.8, ax)
    
    # 添加文本标签
    ax.text(
        x=0.5, y=0.5, s=label, color=colors[0], va="center", ha="center",
        ma="center", fontfamily="Roboto Mono", fontsize=14, fontweight="bold",
        linespacing=0.87, transform=ax.transAxes
    )

# 调整子图间距
fig.subplots_adjust(wspace=0, hspace=0)

output_21_0

多图汇总

# 首先确定多图布局
fig, axes = plt.subplots(nrows=6, ncols=5, figsize=(15, 17.5), subplot_kw={"projection": "polar"})
fig.patch.set_facecolor("white")
gs = axes[0, 0].get_gridspec() # 获取图像的网格规范（gridspec），用于处理子图相对位置和大小

# 移除前四行五列中（也就是前20个）的所有子图
for i in range(4):
    for j in range(5):
        axes[i, j].remove()

# 在移除的子图的位置添加一个大子图
ax_big = fig.add_subplot(
    gs[:4, :], # 大子图将占据前四行和所有列的位置
    projection="polar" # 用一个大的极坐标图填充
)

# 设置大子图的高宽比为1，使得图像看起来更加整齐
ax_big.set_aspect("equal")

# 调整空间布局
fig.subplots_adjust(
    left=0.04, bottom=0.05, right=0.96, top=0.95, # margins
    wspace=0, hspace=0 # spaces
)

output_23_0

# 然后将大图绘制到上面的布局中
circular_plot(ANGLES, HEIGHTS, COLORS, 1.4, ax_big)

# 文本注释
for idx, row in LABELS_DF.iterrows():
    color=COLORS[row["index"]]
    ax_big.text(
        ANGLES[row["x"]], row["y"], row["label"], ha="right", va="center",
        ma="center", size=13, family="Roboto Mono", weight="bold", color=color
    )

ax_big.text(
    x=0.5, y=0.56, s="********\nCracking\nYour Favorite\nPassword\n********",
    color=GREY60, va="center", ha="center", ma="center", fontfamily="Roboto Mono",
    fontsize=30, fontweight="bold", linespacing=0.95, transform=ax_big.transAxes
)

ax_big.text(
    x=0.5, y=0.44, s="Time it takes to crack the 500 most\ncommon passwords by online guessing.\nSorted by rank and colored by category.",
    color=GREY60, va="center", ha="center", ma="center", fontfamily="Roboto Mono",
    fontsize=12, linespacing=0.87, transform=ax_big.transAxes
)

ax_big.text(
    x=0.5, y=0.37, s="Time is displayed on a logarithmic scale\nwith the rings representing one day,\none week, one month, and one year\n(from inner to outer ring).",
    color=GREY60, va="center", ha="center", ma="center", fontfamily="Roboto Mono",
    fontsize=11, linespacing=0.87, transform=ax_big.transAxes
)
fig

# 最后将多子图绘制在下面的布局中
slices = [(i + 4, j) for i in range(2) for j in range(5)] # 多子图的位置

for category, label, idx_ax in zip(CATEGORIES, LABELS, slices):
    ax = axes[idx_ax]
    idx = df_pw.index[df_pw["category"] == category].tolist()
    
    angles = ANGLES[idx]
    heights = HEIGHTS[idx]
    colors = COLORS[idx]
    
    circular_plot(angles, heights, colors, 0.9, ax)
    
    ax.text(
        x=0.5, y=0.5, s=label, color=colors[0], va="center", ha="center",
        ma="center", fontfamily="Roboto Mono", fontsize=14, fontweight="bold",
        linespacing=0.87, transform=ax.transAxes
    )

fig.text(
    x=0.5, y=0.02, s="Visualization by Cédric Scherer - Data by Knowledge is Beautiful",
    color=GREY60, fontsize=14, ha="center", fontfamily="Roboto Mono", fontweight="bold"
)

fig