-
Notifications
You must be signed in to change notification settings - Fork 6
/
4_gen_tfrecord_binseeker.py
310 lines (279 loc) · 14.8 KB
/
4_gen_tfrecord_binseeker.py
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
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
304
305
306
307
308
309
310
#!/usr/bin/env python
# _*_ coding: utf-8 _*_
import tensorflow as tf
import numpy as np
import csv
import os
import time
import networkx as nx
import numba
import itertools
import config
# =========== global parameters ===========
T = 5 # iteration
N = 2 # embedding_depth
D = 8 # dimensional
P = 64 # embedding_size
B = 10 # mini-batch
lr = 0.0001 # learning_rate
# MAX_SIZE = 0 # record the max number of a function's block
epochs = 10
is_debug = True
#
data_folder = config.FEA_DIR
train_file = config.DATASET_DIR + os.sep + "train"+str(config.TRAIN_DATASET_NUM)+"_["+'_'.join(config.STEP3_PORGRAM_ARR)+"].csv"
test_file = config.DATASET_DIR + os.sep + "test"+str(config.TRAIN_DATASET_NUM)+"_["+'_'.join(config.STEP3_PORGRAM_ARR)+"].csv"
valid_file = config.DATASET_DIR + os.sep + "vaild"+str(config.TRAIN_DATASET_NUM)+"_["+'_'.join(config.STEP3_PORGRAM_ARR)+"].csv"
PREFIX = "_"+str(config.TRAIN_DATASET_NUM)+"_["+'_'.join(config.STEP3_PORGRAM_ARR)+"]"
TRAIN_TFRECORD = config.TFRECORD_VULSEEKER_DIR_DIR + os.sep + "train_"+PREFIX+".tfrecord"
TEST_TFRECORD = config.TFRECORD_VULSEEKER_DIR_DIR + os.sep + "test_"+PREFIX+".tfrecord"
VALID_TFRECORD = config.TFRECORD_VULSEEKER_DIR_DIR + os.sep + "valid_"+PREFIX+".tfrecord"
print TRAIN_TFRECORD
# ==================== load the function pairs list ===================
# 1. load_dataset() load the pairs list for learning, which are
# in train.csv, valid.csv, test.csv .
# 1-1. load_csv_as_pair() process each csv file.
# =====================================================================
def load_dataset():
""" load the pairs list for training, testing, validing
"""
train_pair, train_label = load_csv_as_pair(train_file)
valid_pair, valid_label = load_csv_as_pair(valid_file)
test_pair, test_label = load_csv_as_pair(test_file)
return train_pair, train_label, valid_pair, valid_label, test_pair, test_label
def load_csv_as_pair(pair_label_file):
""" load each csv file, which record the pairs list for learning and its label ( 1 or -1 )
csv file : uid, uid, 1/-1 eg: 1.1.128, 1.4.789, -1
pair_dict = {(uid, uid) : -1/1}
"""
pair_list = []
label_list = []
with open(pair_label_file, "r") as fp:
pair_label = csv.reader(fp)
for line in pair_label:
pair_list.append([line[0], line[1]])
label_list.append(int(line[2]))
return pair_list, label_list
# =============== convert the real data to training data ==============
# 1. construct_learning_dataset() combine the dataset list & real data
# 1-1. generate_adj_matrix_pairs() traversal list and construct all the matrixs
# 1-1-1. convert_graph_to_adj_matrix() process each cfg
# 1-2. generate_features_pair() traversal list and construct all functions' feature map
# =====================================================================
# @numba.jit
def construct_learning_dataset(uid_pair_list):
""" Construct pairs dataset to train the model.
attributes:
adj_matrix_all store each pairs functions' graph info, （i,j)=1 present i--》j, others （i,j)=0
features_all store each pairs functions' feature map
"""
print " start generate adj matrix pairs..."
cfgs_1, cfgs_2, dfgs_1, dfgs_2 = generate_graph_pairs(uid_pair_list)
print " start generate features pairs..."
### !!! record the max number of a function's block
feas_1, feas_2, max_size, num1, num2 = generate_features_pair(uid_pair_list)
return cfgs_1, cfgs_2, dfgs_1, dfgs_2, feas_1, feas_2, num1, num2, max_size
# @numba.jit
def generate_graph_pairs(uid_pair_list):
""" construct all the function pairs' cfg matrix.
"""
cfgs_1 = []
cfgs_2 = []
dfgs_1 = []
dfgs_2 = []
# traversal all the pairs
count = 0
for uid_pair in uid_pair_list:
if is_debug:
count += 1
print " %04d cfg, [ %s , %s ]"%(count, uid_pair[0], uid_pair[1])
adj_matrix_pair = []
# each pair process two function
graph_cfg = nx.read_adjlist(os.path.join(config.FEA_DIR, uid_pair[0]+"_cfg.txt"))
# graph = uid_graph[uid_pair[0]]
# origion_adj_1 = np.array(nx.convert_matrix.to_numpy_matrix(graph, dtype=float))
# origion_adj_1.resize(size, size, refcheck=False)
# adj_matrix_all_1.append(origion_adj_1.tolist())
adj_arr = np.array(nx.convert_matrix.to_numpy_matrix(graph_cfg, dtype=float))
adj_str = adj_arr.astype(np.string_)
cfgs_1.append(",".join(list(itertools.chain.from_iterable(adj_str))))
graph_dfg = nx.read_adjlist(os.path.join(data_folder, uid_pair[0]+"_dfg.txt"))
graph= graph_dfg.copy()
for node in graph.nodes():
if not graph_cfg.has_node(node):
graph_dfg.remove_node(node)
graph_dfg.add_nodes_from(graph_cfg)
adj_arr = np.array(nx.convert_matrix.to_numpy_matrix(graph_dfg, dtype=float))
adj_str = adj_arr.astype(np.string_)
dfgs_1.append(",".join(list(itertools.chain.from_iterable(adj_str))))
graph_cfg = nx.read_adjlist(os.path.join(data_folder, uid_pair[1]+"_cfg.txt"))
# graph = uid_graph[uid_pair[1]]
# origion_adj_2 = np.array(nx.convert_matrix.to_numpy_matrix(graph, dtype=float))
# origion_adj_2.resize(size, size, refcheck=False)
# adj_matrix_all_2.append(origion_adj_2.tolist())
adj_arr = np.array(nx.convert_matrix.to_numpy_matrix(graph_cfg, dtype=float))
adj_str = adj_arr.astype(np.string_)
cfgs_2.append(",".join(list(itertools.chain.from_iterable(adj_str))))
graph_dfg = nx.read_adjlist(os.path.join(data_folder, uid_pair[1]+"_dfg.txt"))
graph= graph_dfg.copy()
for node in graph.nodes():
if not graph_cfg.has_node(node):
graph_dfg.remove_node(node)
graph_dfg.add_nodes_from(graph_cfg)
adj_arr = np.array(nx.convert_matrix.to_numpy_matrix(graph_dfg, dtype=float))
adj_str = adj_arr.astype(np.string_)
dfgs_2.append(",".join(list(itertools.chain.from_iterable(adj_str))))
return cfgs_1, cfgs_2, dfgs_1, dfgs_2
# @numba.jit
def generate_features_pair(uid_pair_list):
""" Construct each function pairs' block feature map.
"""
feas_1 = []
feas_2 = []
num1 = []
num2 = []
node_length = []
# traversal all the pairs
count = 0
for uid_pair in uid_pair_list:
if is_debug:
count += 1
print " %04d feature, [ %s , %s ]"%(count, uid_pair[0], uid_pair[1])
node_vector = []
block_feature_dic={}
with open(os.path.join(config.FEA_DIR, uid_pair[0]+"_fea.csv"), "r") as fp:
for line in csv.reader(fp):
if line[0] == "":
continue
# read every bolck's features
block_feature = [float(x) for x in (line[8:15])]
print line[0],block_feature
# 删除某一列特征
# del block_feature[6]
block_feature_dic.setdefault(str(line[0]), block_feature)
graph_cfg = nx.read_adjlist(os.path.join(data_folder, uid_pair[0] + "_cfg.txt"))
for node in graph_cfg.nodes():
node_vector.append(block_feature_dic[node])
node_length.append(len(node_vector))
num1.append(len(node_vector))
node_arr = np.array(node_vector)
node_str = node_arr.astype(np.string_)
feas_1.append(",".join(list(itertools.chain.from_iterable(node_str))))
node_vector = []
block_feature_dic={}
with open(os.path.join(data_folder, uid_pair[1]+"_fea.csv"), "r") as fp:
for line in csv.reader(fp):
if line[0] == "":
continue
# read every bolck's features
block_feature = [float(x) for x in (line[8:15])]
# 删除某一列特征
# del block_feature[6]
block_feature_dic.setdefault(str(line[0]), block_feature)
graph_cfg = nx.read_adjlist(os.path.join(data_folder, uid_pair[1] + "_cfg.txt"))
for node in graph_cfg.nodes():
node_vector.append(block_feature_dic[node])
node_length.append(len(node_vector))
num2.append(len(node_vector))
node_arr = np.array(node_vector)
node_str = node_arr.astype(np.string_)
feas_2.append(",".join(list(itertools.chain.from_iterable(node_str))))
num1_re = np.array(num1)
num2_re = np.array(num2)
#num1_re = num1_arr.astype(np.string_)
#num2_re = num2_arr.astype(np.string_)
return feas_1, feas_2, np.max(node_length),num1_re,num2_re
# ========================== the main function ========================
# 1. load_dataset() load the train, valid, test csv file.
# 2. load_all_data() load the origion data, including block info, cfg by networkx.
# 3. construct_learning_dataset() combine the csv file and real data, construct training dataset.
# =====================================================================
# 1. load the train, valid, test csv file.
data_time = time.time()
train_pair, train_label, valid_pair, valid_label, test_pair, test_label = load_dataset()
print "1. loading pairs list time", time.time() - data_time, "(s)"
# 2. load the origion data, including block info, cfg by networkx.
# graph_time = time.time()
# uid_cfg,uid_dfg, fea_dict = load_all_data()
# print "2. loading graph data time", time.time() - graph_time, "(s)"
# 3. construct training dataset.
cons_time = time.time()
# ======================= construct train data =====================
train_cfg_1, train_cfg_2,train_dfg_1, train_dfg_2,train_fea_1, train_fea_2, train_num1, train_num2, train_max \
= construct_learning_dataset(train_pair)
# ========================== store in pickle ========================
node_list = np.linspace(train_max,train_max, len(train_label),dtype=int)
writer = tf.python_io.TFRecordWriter(TRAIN_TFRECORD)
for item1,item2,item3,item4,item5,item6, item7, item8, item9, item10 in itertools.izip(
train_label, train_cfg_1, train_cfg_2, train_dfg_1, train_dfg_2, train_fea_1, train_fea_2,
train_num1, train_num2, node_list):
example = tf.train.Example(
features = tf.train.Features(
feature = {
'label':tf.train.Feature(int64_list = tf.train.Int64List(value=[item1])),
'cfg_1': tf.train.Feature(bytes_list=tf.train.BytesList(value=[item2])),
'cfg_2': tf.train.Feature(bytes_list=tf.train.BytesList(value=[item3])),
'dfg_1': tf.train.Feature(bytes_list=tf.train.BytesList(value=[item4])),
'dfg_2': tf.train.Feature(bytes_list=tf.train.BytesList(value=[item5])),
'fea_1': tf.train.Feature(bytes_list=tf.train.BytesList(value=[item6])),
'fea_2': tf.train.Feature(bytes_list=tf.train.BytesList(value=[item7])),
'num1':tf.train.Feature(int64_list = tf.train.Int64List(value=[item8])),
'num2':tf.train.Feature(int64_list = tf.train.Int64List(value=[item9])),
'max': tf.train.Feature(int64_list=tf.train.Int64List(value=[item10]))}))
serialized = example.SerializeToString()
writer.write(serialized)
writer.close()
# ========================== clean memory ========================
# del train_pair, train_adj_matrix_1,train_adj_matrix_2,train_feature_map_1,train_feature_map_2,train_max
# ======================= construct valid data =====================
valid_cfg_1, valid_cfg_2, valid_dfg_1, valid_dfg_2, valid_fea_1, valid_fea_2, valid_num1, valid_num2, valid_max \
= construct_learning_dataset(valid_pair)
# ========================== store in pickle ========================
node_list = np.linspace(valid_max,valid_max, len(valid_label),dtype=int)
writer = tf.python_io.TFRecordWriter(VALID_TFRECORD)
for item1,item2,item3,item4,item5,item6, item7, item8, item9, item10 in itertools.izip(
valid_label, valid_cfg_1, valid_cfg_2, valid_dfg_1, valid_dfg_2, valid_fea_1, valid_fea_2,
valid_num1, valid_num2, node_list):
example = tf.train.Example(
features = tf.train.Features(
feature = {
'label':tf.train.Feature(int64_list = tf.train.Int64List(value=[item1])),
'cfg_1': tf.train.Feature(bytes_list=tf.train.BytesList(value=[item2])),
'cfg_2': tf.train.Feature(bytes_list=tf.train.BytesList(value=[item3])),
'dfg_1': tf.train.Feature(bytes_list=tf.train.BytesList(value=[item4])),
'dfg_2': tf.train.Feature(bytes_list=tf.train.BytesList(value=[item5])),
'fea_1': tf.train.Feature(bytes_list=tf.train.BytesList(value=[item6])),
'fea_2': tf.train.Feature(bytes_list=tf.train.BytesList(value=[item7])),
'num1':tf.train.Feature(int64_list = tf.train.Int64List(value=[item8])),
'num2':tf.train.Feature(int64_list = tf.train.Int64List(value=[item9])),
'max': tf.train.Feature(int64_list=tf.train.Int64List(value=[item10]))}))
serialized = example.SerializeToString()
writer.write(serialized)
writer.close()
# ========================== clean memory ========================
# del valid_pair, valid_adj_matrix_1,valid_adj_matrix_2,valid_feature_map_1,valid_feature_map_2,valid_max
# ======================= construct test data =====================
test_cfg_1, test_cfg_2, test_dfg_1, test_dfg_2, test_fea_1, test_fea_2,test_num1, test_num2, test_max \
= construct_learning_dataset(test_pair)
# ========================== store in pickle ========================
node_list = np.linspace(test_max,test_max, len(test_label),dtype=int)
writer = tf.python_io.TFRecordWriter(TEST_TFRECORD)
for item1,item2,item3,item4,item5,item6, item7, item8, item9, item10 in itertools.izip(
test_label, test_cfg_1, test_cfg_2, test_dfg_1, test_dfg_2, test_fea_1, test_fea_2,
test_num1, test_num2, node_list):
example = tf.train.Example(
features = tf.train.Features(
feature = {
'label':tf.train.Feature(int64_list = tf.train.Int64List(value=[item1])),
'cfg_1': tf.train.Feature(bytes_list=tf.train.BytesList(value=[item2])),
'cfg_2': tf.train.Feature(bytes_list=tf.train.BytesList(value=[item3])),
'dfg_1': tf.train.Feature(bytes_list=tf.train.BytesList(value=[item4])),
'dfg_2': tf.train.Feature(bytes_list=tf.train.BytesList(value=[item5])),
'fea_1': tf.train.Feature(bytes_list=tf.train.BytesList(value=[item6])),
'fea_2': tf.train.Feature(bytes_list=tf.train.BytesList(value=[item7])),
'num1':tf.train.Feature(int64_list = tf.train.Int64List(value=[item8])),
'num2':tf.train.Feature(int64_list = tf.train.Int64List(value=[item9])),
'max': tf.train.Feature(int64_list=tf.train.Int64List(value=[item10]))}))
serialized = example.SerializeToString()
writer.write(serialized)
writer.close()