EUF/plot_pool.py at master · OSUSecLab/EUF · GitHub

1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
#!/usr/bin/env python3
"""
Plot pool
"""
importglob
importnumpyasnp
importpandasaspd
importmatplotlib.pyplotasplt


PROFILE_ORDER= {"Alice": 0, "Bob": 1, "Charlie": 2, "Diane": 3}
PROFILE_TO_LETTER= {"Alice": "A", "Bob": "B", "Charlie": "C", "Diane": "D"}

PROFILE_COLORS= {
"Alice": "steelblue",
"Bob": "darkorange",
"Charlie": "seagreen",
"Diane": "mediumpurple",
}

PROFILE_LABELS= {
"Alice": "Deliberative",
"Bob": "Low Engagement",
"Charlie": "High Frequency",
"Diane": "Intermittent",
}


deftrim_min_max_by_entropy(group: pd.DataFrame) ->pd.DataFrame:
iflen(group) <=2:
returngroup.copy()

g=group.sort_values("entropy_loss_bits", kind="mergesort").reset_index(drop=True)
# drop first and last
returng.iloc[1:-1].copy()


# X tick labels: A_i, B_i, C_i, D_i
defto_letter_subscript(t: str) ->str:
p, idx=t.split("_")
letter=PROFILE_TO_LETTER.get(p, p[0].upper())

returnf"${letter}_{{{idx}}}$"


defmain():
files=sorted(glob.glob("data/qif_pool_results_run*.csv"))
df=pd.concat([pd.read_csv(f) forfinfiles], ignore_index=True)

df["profile"] =df["target"].apply(lambdax: x.split("_")[0])
df["index"] =df["target"].apply(lambdax: int(x.split("_")[1]))
df["sort_key"] =df["profile"].map(PROFILE_ORDER).fillna(99) +df["index"] /100.0

# allow for various pool size
N_pool=int(df["target"].nunique())
vuln_baseline=1.0/N_pool

# trim outliers
trimmed= (df.groupby("target", group_keys=False).apply(trim_min_max_by_entropy).reset_index(drop=True))

# per user
agg=trimmed.groupby("target").agg(
loss_mean=("entropy_loss_bits", "mean"),
loss_std=("entropy_loss_bits", "std"),
vuln_mean=("bayes_vuln", "mean"),
vuln_std=("bayes_vuln", "std"),
profile=("profile", "first"),
index=("index", "first"),
n_used=("entropy_loss_bits", "count"),
 ).reset_index()

agg["sort_key"] =agg["profile"].map(PROFILE_ORDER).fillna(99) +agg["index"] /100.0
agg=agg.sort_values("sort_key").reset_index(drop=True)

agg["color"] =agg["profile"].map(PROFILE_COLORS).fillna("gray")


xtick_labels=agg["target"].apply(to_letter_subscript).tolist()
x=np.arange(len(agg))

fig, ax1=plt.subplots(figsize=(10, 5))

bars=ax1.bar(
x,
agg["loss_mean"].values,
yerr=agg["loss_std"].fillna(0.0).values,
capsize=4,
color=agg["color"].values,
alpha=0.95,
error_kw={"ecolor": "tab:blue", "elinewidth": 1.5},
 )
ax1.set_ylabel("Entropy Loss ($\\Delta H$, bits)")
ax1.set_xticks(x)
ax1.set_xticklabels(xtick_labels, rotation=25, ha="center")

# baseline
ax1.axhline(y=0.0, color="gray", linestyle="--", linewidth=1)

# BV
ax2=ax1.twinx()
ax2.errorbar(
x,
agg["vuln_mean"].values,
yerr=agg["vuln_std"].fillna(0.0).values,
fmt="o",
capsize=4,
color="black",
ecolor="black",
elinewidth=1.2,
markersize=5,
 )
ax2.set_ylabel("Bayes Vulnerability ($V$)")

# baseline uniform prior
ax2.axhline(y=vuln_baseline, color="black", linestyle=":", linewidth=1)

ax1.set_title(f"Pool Evaluation (N={N_pool}): $\\Delta H$ and Bayes Vulnerability")

# legend
handles= []
seen=set()
forpinagg["profile"].unique():
label=PROFILE_LABELS.get(p, p)
iflabelnotinseen:
handles.append(plt.Line2D([], [], color=PROFILE_COLORS.get(p, "gray"), marker="s", linestyle="None", label=label))
seen.add(label)

handles.append(plt.Line2D([], [], color="black", marker="o", linestyle="None", label="Bayes Vulnerability"))
handles.append(plt.Line2D([], [], color="gray", linestyle="--", label="$\\Delta H$ baseline (0)"))
handles.append(plt.Line2D([], [], color="black", linestyle=":", label=f"$V$ baseline (1/{N_pool})"))

ax1.legend(handles=handles, loc="upper left", fontsize=9)

ax1.grid(axis="y", linestyle="--", alpha=0.5)

plt.tight_layout()
out="out/pool_entropy_vulnerability.pdf"
plt.savefig(out)
plt.show()

# Summary stats
print("\n==== Pool Summary ====")
for_, rinagg.iterrows():
ls=0.0ifpd.isna(r["loss_std"]) elser["loss_std"]
vs=0.0ifpd.isna(r["vuln_std"]) elser["vuln_std"]
print(f"{r['target']}: n={int(r['n_used'])} ΔH={r['loss_mean']:.4f}+-{ls:.4f} V={r['vuln_mean']:.4f}+-{vs:.4f}")

print(f"Created {out}")


if__name__=="__main__":
main()