-
Notifications
You must be signed in to change notification settings - Fork 0
/
Utils.py
220 lines (182 loc) · 9.31 KB
/
Utils.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
import numpy as np
import pandas as pd
from sklearn.metrics import f1_score
from collections import defaultdict
from Classifiers import *
from time import time
import pickle, os
from imblearn.under_sampling import RandomUnderSampler
from imblearn.over_sampling import SVMSMOTE, ADASYN, BorderlineSMOTE
from imblearn.combine import SMOTEENN
from tabulate import tabulate
from sklearn.preprocessing import normalize
from matplotlib import pyplot as plt
def compare_classifiers(x_old, y_old, x_test, y_test, data_generator, label_mapping, models,cv=3):
"""Compares the perfomace of models using recall, precision & fscore. Dumps the results in .txt file"""
a = {'Analysis':6000, 'Backdoor':6000, 'DoS':3000, 'Exploits':0,'Fuzzers':2000,\
'Generic':0, 'Reconnaissance':5000, 'Shellcode':2000, 'Worms':300}
temp = np.array(list(a.values()))
p = temp/temp.sum()
perfomace_results = defaultdict(defaultdict(dict).copy)
val , count = np.unique(y_old,return_counts =True)
current_gt = dict(zip(val , count))
up_sampling_strategy = {label_mapping.get(j) : i+current_gt.get(label_mapping.get(j)) for j, i in a.items() if i > 0}
# up_sampling_strategy = {}
print(up_sampling_strategy)
start_t = time()
nameUpsampler = "DGM"
if isinstance(data_generator,str):
print(f'Using : {data_generator}')
up_sampling_strategy = {label_mapping["probe"]:14656, label_mapping["r2l"]:13995,label_mapping["u2r"]:10052}
if data_generator == "ADASYN":
sm = ADASYN(sampling_strategy = up_sampling_strategy,n_jobs=-1)
elif data_generator == "SMOTEENN":
sm = SMOTEENN(sampling_strategy = up_sampling_strategy,n_jobs=-1)
elif data_generator == "BorderlineSMOTE" :
sm = BorderlineSMOTE(sampling_strategy = up_sampling_strategy,n_jobs=-1)
else:
sm = SVMSMOTE(sampling_strategy = up_sampling_strategy,svm_estimator=SVC(C=10, cache_size=1500, class_weight='balanced'))
sm.fit(x_old,y_old)
nameUpsampler = type(sm).__name__
elapsed_time = time() - start_t
print(f"Time taken : {elapsed_time}")
rus = RandomUnderSampler(sampling_strategy = {label_mapping["Generic"]:20000,label_mapping["Exploits"]:20000},random_state=42)
x_old , y_old = rus.fit_resample(x_old,y_old)
for i in range(cv):
print(f"Cross validation number : {i+1}")
if not isinstance(data_generator,str):
labels = np.random.choice(list(label_mapping.values()),(temp.sum(),1),p=p,replace=True)
rand_noise_dim = data_generator.input_shape[0][-1]
noise = np.random.normal(0, 1, (len(labels), rand_noise_dim))
#print([noise, labels])
generated_x = normalize_data(data_generator.predict([noise, labels])[:,:-1],None)
new_trainx = np.vstack([x_old,generated_x])
new_y = np.append(y_old,labels)
else :
start_t = time()
sub_x, sub_y = subsample(x=x_old,y=y_old,size=25000)
new_trainx, new_y = sm.fit_resample(sub_x, sub_y)
elapsed_time = time() - start_t
print(f"Time taken : {elapsed_time}")
randf = random_forest(new_trainx, new_y, x_test, y_test,label_mapping)
nn = neural_network(new_trainx, new_y, x_test, y_test,label_mapping,True)
deci = decision_tree(new_trainx, new_y, x_test, y_test,label_mapping)
sVmclf = svm(new_trainx, new_y, x_test, y_test,label_mapping,True)
for estimator in [randf,deci,nn,sVmclf] :
name = estimator.__class__.__name__
pred = estimator.predict(x_test)
precision,recall,fscore,_ = precision_recall_fscore_support(y_test,pred,labels=list(label_mapping.values()))
perfomace_results[name][i]["precision"] = precision.tolist()
perfomace_results[name][i]["recall"] = recall.tolist()
perfomace_results[name][i]["fscore"] = fscore.tolist()
perfomace_results[name][i]["weighted_f1"] = [f1_score(y_test,pred,labels=list(label_mapping.values()),average='weighted')] * len(label_mapping)
t = int(time())
for estimator in perfomace_results.keys():
tempdf = pd.DataFrame.from_dict(perfomace_results[estimator][0])
tempdf.index = list(label_mapping.values())
for i in range(1,cv):
to_append = pd.DataFrame.from_dict(perfomace_results[estimator][i])
to_append.index = list(label_mapping.values())
tempdf = tempdf.append(to_append)
tempdf["class"] = tempdf.index.map({y:x for x,y in label_mapping.items()})
tempdf = tempdf.groupby("class").agg({'precision': ['mean', 'std'], 'recall': ['mean', 'std'], 'fscore': ['mean', 'std'], 'weighted_f1' : ['mean', 'std']})
tempdf.columns = [f"{i[0]}_{i[1]}" for i in tempdf.columns]
with open(f'./results/performance_{nameUpsampler}_{cv}validations.txt', 'a') as outputfile:
outputfile.write("\n"+estimator+"\n")
print(tabulate(tempdf, headers='keys', tablefmt='psql'), file=outputfile)
with open(f'./results/performance_{nameUpsampler}_{cv}validations.txt', 'a') as outputfile:
outputfile.write(nameUpsampler)
def save_classifiers(clfs, dir = "./trained_classifiers"):
if not os.path.exists(dir):
os.makedirs(dir)
for classifier in clfs:
with open(os.path.join(dir ,f"{classifier.__class__.__name__}.pickle"), "wb") as f:
pickle.dump(classifier,f)
print("classifiers Save : [DONE]")
def load_pretrained_classifiers(dir = "./trained_classifiers"):
assert os.path.exists(dir), f'{dir} : [does not exists]'
res = {}
for file in os.listdir(dir):
file_full_path = os.path.join(dir, file)
if os.path.isfile(file_full_path) and file_full_path[-7:] == ".pickle":
with open(file_full_path,"rb") as f:
clf = pickle.load(f)
res[clf.__class__.__name__] = clf
if len(res) > 0: print("pretrained classifiers load : [DONE]")
return res
def normalize_data(X,data_cols):
"""Scale input vectors individually to unit norm (vector length)"""
if data_cols is None:
return normalize(X)
else :
X[data_cols] = normalize(X[data_cols])
return X
def subsample(x, y, size=25000):
"""Sample data from x """
new_counts = np.ceil(np.bincount(y)/y.shape[0] * size)
data = np.hstack((x, y.reshape((y.shape[0], 1))))
class_wise_arrays = []
for cls in range(10):
class_wise_arrays.append(data[y == cls])
class_wise_random_indices = []
for array, count in zip(class_wise_arrays, new_counts):
class_wise_random_indices.append(np.random.choice(array.shape[0], int(count), replace=False))
data_small = np.vstack(tuple([array[index, :] for array, index in zip(class_wise_arrays, class_wise_random_indices)]))
x_small = data_small[:, :-1]
y_small = data_small[:, -1]
return x_small, y_small.astype("int")
def plot_training_summary(filePath='',savePath='.s/imgs'):
"""
Plot and Save GAN training metrices i.e discriminator, generator loss & accuracy, KL-Divergence
:param filePath : sting, path to file contatining GAN training logs file
:param savePath : string, directory path to save resuting plot
:return : None
"""
assert os.path.isfile(filePath) , f'{filePath} does not exist'
try:
with open(filePath, 'rb') as f:
x = pickle.load(f)
except :
print(f'could not open {filePath}')
exit(1)
d_l = np.array(x['discriminator_loss']).ravel()
g_l = np.array(x['Generator_loss']).ravel()
acc_history = np.array(x['acc_history'])
acc = acc_history.sum(axis=1) * 0.5
acc_real = acc_history[:,1]
acc_gen = acc_history[:,0]
kl = np.array(x["kl_divergence"]).ravel()
n = np.arange(len(d_l))
figname = os.path.split(filePath)[1].replace('.pickle','')
title = 'Loss and Accuracy plot'+'\n'+ figname
fig = plt.figure(figsize=(19.20,10.80))
fig.suptitle(title, fontsize=15,fontweight="bold")
axs1 = plt.subplot(222)
# axs1.set_title(title,fontsize=5.0,fontweight="bold")
axs1.plot(n, g_l,label='Generator loss',linewidth=3)
axs1.plot(n, d_l,label='Discriminator loss',linewidth=3)
axs1.legend(loc=0, prop={'size': 13})
axs1.set_ylabel('Loss',fontsize=15.0,fontweight="bold")
axs1.tick_params(labelsize=10)
# axs1.tick_params(axis='x',which='both',bottom=False,top=False,labelbottom=False,labelsize=20)
# axs2.plot(n, acc,'r',label='Discriminator accuracy',linewidth=4)
axs2 = plt.subplot(221)
axs2.plot(n, acc_gen,label='Accuracy on Generated',linewidth=3)
axs2.plot(n, acc_real,label='Accuracy on Real',linewidth=3)
axs2.legend(loc=0,prop={'size': 13})
axs2.set_ylabel('Accuracy',fontsize=15.0,fontweight="bold")
# axs2.set_xlabel('Epoch',fontsize=15.0,fontweight="bold")
axs2.tick_params(labelsize=10)
axs3 = plt.subplot(212)
n = np.arange(0,(len(kl)*10),10)
axs3.plot(n, kl,label='KL',linewidth=3)
axs3.legend(loc=0,prop={'size': 13})
axs3.set_ylabel('KL-Divergence',fontsize=15.0,fontweight="bold")
axs3.set_xlabel('Epoch',fontsize=15.0,fontweight="bold")
axs3.tick_params(labelsize=10)
# plt.tight_layout()
fig.tight_layout(rect=[0, 0.03, 1, 0.95])
if not os.path.exists(savePath):
os.makedirs(savePath)
plt.savefig(os.path.join(savePath,figname+'.png'),dpi = 300)
plt.close('all') #plt.close(fig)