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
|
from utils import *
from os import listdir
import argparse
from scipy.optimize import brentq
from scipy.interpolate import interp1d
def get_localization_acc(adv_pkt_idx_dict, top_loss_idx_dict, ngram=3):
def ngram_idx_lst(lst, ngram):
new_lst = []
for top_loss_idx in lst:
new_lst.append(top_loss_idx)
for delta in range(1, ngram):
new_lst.append(top_loss_idx + delta)
return new_lst
hit_dict = {}
top_1_hit_cnt = 0
top_3_hit_cnt = 0
top_5_hit_cnt = 0
for conn_id, adv_pkt_idx_lst in adv_pkt_idx_dict.items():
if conn_id not in top_loss_idx_dict:
continue
top_loss_idx_lst = top_loss_idx_dict[conn_id]
if len(top_loss_idx_lst) == 0:
continue
adv_pkt_idx_set = set(adv_pkt_idx_lst)
top_1 = ngram_idx_lst([top_loss_idx_lst[0]], ngram)
if len(top_loss_idx_lst) >= 5:
top_5 = ngram_idx_lst(top_loss_idx_lst[:5], ngram)
else:
top_5 = ngram_idx_lst(top_loss_idx_lst, ngram)
if len(top_loss_idx_lst) >= 3:
top_3 = ngram_idx_lst(top_loss_idx_lst[:3], ngram)
else:
top_3 = ngram_idx_lst(top_loss_idx_lst, ngram)
top_1_set = set(top_1)
top_3_set = set(top_3)
top_5_set = set(top_5)
if len(top_1_set.intersection(adv_pkt_idx_set)) != 0:
top_1_hit = True
else:
top_1_hit = False
if len(top_3_set.intersection(adv_pkt_idx_set)) != 0:
top_3_hit = True
else:
top_3_hit = False
if len(top_5_set.intersection(adv_pkt_idx_set)) != 0:
top_5_hit = True
else:
top_5_hit = False
hit_dict[conn_id] = (top_1_hit, top_3_hit, top_5_hit)
for conn_id, hits in hit_dict.items():
if hits[0]:
top_1_hit_cnt += 1
if hits[1]:
top_3_hit_cnt += 1
if hits[2]:
top_5_hit_cnt += 1
top_1_hit_acc = top_1_hit_cnt / len(hit_dict)
top_3_hit_acc = top_3_hit_cnt / len(hit_dict)
top_5_hit_acc = top_5_hit_cnt / len(hit_dict)
return top_1_hit_acc, top_3_hit_acc, top_5_hit_acc
if __name__ == "__main__":
parser = argparse.ArgumentParser(
description='Let us compute the ROC curve.')
parser.add_argument('--loss-list-fname', type=str,
help='1st path to loss list')
parser.add_argument('--n-gram', type=int)
parser.add_argument('--attack-info-fpath', type=str,
help='to calculate the localization perf')
parser.add_argument('--loss-list-dir', type=str,
help='1st path to loss list')
parser.add_argument('--fig-fpath', type=str, help='path to save figure')
parser.add_argument('--ds-title', type=str,
default='SymTCP', help='title in curve painting')
parser.add_argument('--balance', action='store_true',
default=False, help='whether to balance by labels')
args = parser.parse_args()
all_losslist_files = listdir(args.loss_list_dir)
losslist_files = {}
for fname in all_losslist_files:
if args.loss_list_fname in fname:
label = fname.split('.')[-1]
attack_type = fname.split('.')[-9]
losslist_files[label] = (fname, attack_type)
adv_pkt_idx_dict = {}
with open(args.attack_info_fpath, 'r') as fin:
data = fin.readlines()
for row in data:
row = row.rstrip('\n')
conn_id, adv_pkt_info = row.split('\t')
adv_pkt_info = eval(adv_pkt_info)
adv_pkt_idx_dict[conn_id] = adv_pkt_info
for label, (fname, attack_type) in losslist_files.items():
lossfile_fname = '/'.join([args.loss_list_dir, fname])
y, scores, top_loss_lst = read_loss_list(
lossfile_fname, balance_by_label=args.balance)
if len(y) == 0:
print(">>>>>>>>> [Label %s; Attack type %s] Not found <<<<<<<<<" % (
label, attack_type))
continue
#
top_1_hit_acc, top_3_hit_acc, top_5_hit_acc = get_localization_acc(
adv_pkt_idx_dict, top_loss_lst, ngram=args.n_gram)
y = np.array(y)
scores = np.array(scores)
fpr, tpr, thresholds = metrics.roc_curve(y, scores, pos_label=1)
tpr_lst = {}
for fpr_th in [0.005,0.07,0.09]:
for i in range(1, len(fpr)):
curr_fpr = fpr[i]
prev_fpr = fpr[i-1]
if prev_fpr <= fpr_th and fpr_th <= curr_fpr:
tpr_lst[fpr_th] = tpr[i-1]
eer_score = brentq(lambda x: 1. - x - interp1d(fpr, tpr)(x), 0., 1.)
score = metrics.roc_auc_score(y, scores)
print(','.join([label, attack_type, str(score), str(tpr_lst[0.005]),str(
tpr_lst[0.07]), str(tpr_lst[0.09]), str(eer_score), str(top_1_hit_acc), str(top_3_hit_acc), str(top_5_hit_acc)]))
|
