Newer
Older
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
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
# -*- coding: utf-8 -*-
import pandas as pd
import numpy as np
import torch
import matplotlib.pyplot as plt
from torch.utils.data import DataLoader, random_split, Subset
from sklearn.preprocessing import MinMaxScaler
from util.env import get_device, set_device
from util.preprocess import build_loc_net, construct_data
from util.net_struct import get_feature_map, get_fc_graph_struc
from util.iostream import printsep
from datasets.TimeDataset import TimeDataset
from models.GDN import GDN
from train import train
from test import test
from evaluate import get_err_scores, get_best_performance_data, get_val_performance_data, get_full_err_scores
import sys
from datetime import datetime
import os
import argparse
from pathlib import Path
import matplotlib.pyplot as plt
import json
import random
class Main():
def __init__(self, train_config, env_config, debug=False):
self.train_config = train_config
self.env_config = env_config
self.datestr = None
dataset = self.env_config['dataset']
train_orig = pd.read_csv(f'./data/{dataset}/train.csv', sep=',', index_col=0)
test_orig = pd.read_csv(f'./data/{dataset}/test.csv', sep=',', index_col=0)
train, test = train_orig, test_orig
if 'attack' in train.columns:
train = train.drop(columns=['attack'])
feature_map = get_feature_map(dataset)
fc_struc = get_fc_graph_struc(dataset)
set_device(env_config['device'])
self.device = get_device()
fc_edge_index = build_loc_net(fc_struc, list(train.columns), feature_map=feature_map)
fc_edge_index = torch.tensor(fc_edge_index, dtype = torch.long)
self.feature_map = feature_map
train_dataset_indata = construct_data(train, feature_map, labels=0)
test_dataset_indata = construct_data(test, feature_map, labels=test.attack.tolist())
cfg = {
'slide_win': train_config['slide_win'],
'slide_stride': train_config['slide_stride'],
}
train_dataset = TimeDataset(train_dataset_indata, fc_edge_index, mode='train', config=cfg)
test_dataset = TimeDataset(test_dataset_indata, fc_edge_index, mode='test', config=cfg)
train_dataloader, val_dataloader = self.get_loaders(train_dataset, train_config['seed'], train_config['batch'], val_ratio = train_config['val_ratio'])
self.train_dataset = train_dataset
self.test_dataset = test_dataset
self.train_dataloader = train_dataloader
self.val_dataloader = val_dataloader
self.test_dataloader = DataLoader(test_dataset, batch_size=train_config['batch'],
shuffle=False, num_workers=0)
edge_index_sets = []
edge_index_sets.append(fc_edge_index)
self.model = GDN(edge_index_sets, len(feature_map),
dim=train_config['dim'],
input_dim=train_config['slide_win'],
out_layer_num=train_config['out_layer_num'],
out_layer_inter_dim=train_config['out_layer_inter_dim'],
topk=train_config['topk']
).to(self.device)
def run(self):
if len(self.env_config['load_model_path']) > 0:
model_save_path = self.env_config['load_model_path']
else:
model_save_path = self.get_save_path()[0]
self.train_log = train(self.model, model_save_path,
config = train_config,
train_dataloader=self.train_dataloader,
val_dataloader=self.val_dataloader,
feature_map=self.feature_map,
test_dataloader=self.test_dataloader,
test_dataset=self.test_dataset,
train_dataset=self.train_dataset,
dataset_name=self.env_config['dataset']
)
# test
self.model.load_state_dict(torch.load(model_save_path))
best_model = self.model.to(self.device)
_, self.test_result = test(best_model, self.test_dataloader)
_, self.val_result = test(best_model, self.val_dataloader)
self.get_score(self.test_result, self.val_result)
def get_loaders(self, train_dataset, seed, batch, val_ratio=0.1):
dataset_len = int(len(train_dataset))
train_use_len = int(dataset_len * (1 - val_ratio))
val_use_len = int(dataset_len * val_ratio)
val_start_index = random.randrange(train_use_len)
indices = torch.arange(dataset_len)
train_sub_indices = torch.cat([indices[:val_start_index], indices[val_start_index+val_use_len:]])
train_subset = Subset(train_dataset, train_sub_indices)
val_sub_indices = indices[val_start_index:val_start_index+val_use_len]
val_subset = Subset(train_dataset, val_sub_indices)
train_dataloader = DataLoader(train_subset, batch_size=batch,
shuffle=True)
val_dataloader = DataLoader(val_subset, batch_size=batch,
shuffle=False)
return train_dataloader, val_dataloader
def get_score(self, test_result, val_result):
feature_num = len(test_result[0][0])
np_test_result = np.array(test_result)
np_val_result = np.array(val_result)
test_labels = np_test_result[2, :, 0].tolist()
test_scores, normal_scores = get_full_err_scores(test_result, val_result)
top1_best_info = get_best_performance_data(test_scores, test_labels, topk=1)
top1_val_info = get_val_performance_data(test_scores, normal_scores, test_labels, topk=1)
print('=========================** Result **============================\n')
info = None
if self.env_config['report'] == 'best':
info = top1_best_info
elif self.env_config['report'] == 'val':
info = top1_val_info
print(f'F1 score: {info[0]}')
print(f'precision: {info[1]}')
print(f'recall: {info[2]}\n')
def get_save_path(self, feature_name=''):
dir_path = self.env_config['save_path']
if self.datestr is None:
now = datetime.now()
self.datestr = now.strftime('%m|%d-%H:%M:%S')
datestr = self.datestr
paths = [
f'./pretrained/{dir_path}/best_{datestr}.pt',
f'./results/{dir_path}/{datestr}.csv',
]
for path in paths:
dirname = os.path.dirname(path)
Path(dirname).mkdir(parents=True, exist_ok=True)
return paths
if __name__ == "__main__":
parser = argparse.ArgumentParser()
parser.add_argument('-batch', help='batch size', type = int, default=128)
parser.add_argument('-epoch', help='train epoch', type = int, default=100)
parser.add_argument('-slide_win', help='slide_win', type = int, default=15)
parser.add_argument('-dim', help='dimension', type = int, default=64)
parser.add_argument('-slide_stride', help='slide_stride', type = int, default=5)
parser.add_argument('-save_path_pattern', help='save path pattern', type = str, default='')
parser.add_argument('-dataset', help='wadi / swat', type = str, default='wadi')
parser.add_argument('-device', help='cuda / cpu', type = str, default='cuda')
parser.add_argument('-random_seed', help='random seed', type = int, default=0)
parser.add_argument('-comment', help='experiment comment', type = str, default='')
parser.add_argument('-out_layer_num', help='outlayer num', type = int, default=1)
parser.add_argument('-out_layer_inter_dim', help='out_layer_inter_dim', type = int, default=256)
parser.add_argument('-decay', help='decay', type = float, default=0)
parser.add_argument('-val_ratio', help='val ratio', type = float, default=0.1)
parser.add_argument('-topk', help='topk num', type = int, default=20)
parser.add_argument('-report', help='best / val', type = str, default='best')
parser.add_argument('-load_model_path', help='trained model path', type = str, default='')
args = parser.parse_args()
random.seed(args.random_seed)
np.random.seed(args.random_seed)
torch.manual_seed(args.random_seed)
torch.cuda.manual_seed(args.random_seed)
torch.cuda.manual_seed_all(args.random_seed)
torch.backends.cudnn.benchmark = False
torch.backends.cudnn.deterministic = True
os.environ['PYTHONHASHSEED'] = str(args.random_seed)
train_config = {
'batch': args.batch,
'epoch': args.epoch,
'slide_win': args.slide_win,
'dim': args.dim,
'slide_stride': args.slide_stride,
'comment': args.comment,
'seed': args.random_seed,
'out_layer_num': args.out_layer_num,
'out_layer_inter_dim': args.out_layer_inter_dim,
'decay': args.decay,
'val_ratio': args.val_ratio,
'topk': args.topk,
}
env_config={
'save_path': args.save_path_pattern,
'dataset': args.dataset,
'report': args.report,
'device': args.device,
'load_model_path': args.load_model_path
}
main = Main(train_config, env_config, debug=False)
main.run()