-
Notifications
You must be signed in to change notification settings - Fork 0
/
dns_analytics.py
214 lines (163 loc) · 8.1 KB
/
dns_analytics.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
from functools import total_ordering
import networkx
import pandas as pd
import numpy as np
from scipy.stats import kruskal
from sklearn.cluster import AgglomerativeClustering
import argparse
import os
import us
from pylab import plot, show
import matplotlib.pyplot as plt
import matplotlib.colors as mcolors
def main():
parser = argparse.ArgumentParser()
parser.add_argument("-z", default=2, type=int, help="Maximum z score for a pair of servers. Default is 2")
parser.add_argument("-v", default=0.1, type=float, help="Maximum coefficient of variation for a point of measurement between a pair of servers. Default is 0.1")
parser.add_argument("-c", default="rtt", type=str, help="rtt or rtt_normalized. Default is rtt")
parser.add_argument("-w", default=None, type=str, help="Path to a weights file. Default is None")
args = parser.parse_args()
column = args.c
pickle_filename = get_or_create_pickle(column=column, cov_max=args.v, z=args.z)
print("Generating results...")
pairs = pd.read_pickle(pickle_filename)
if args.w is not None:
if os.path.isfile(args.w):
weights = pd.read_csv(args.w)
weights['authoritative_state'] = weights['state'].map(us.states.mapping('name', 'abbr'))
weights = weights[['authoritative_state', 'value']]
weights.value = weights.value.astype(int)
pairs = pairs.reset_index().merge(weights, how='inner')
pairs['median'] = pairs['median'] * pairs['value']
del pairs['value']
pairs = pairs.set_index(['recursive_state', 'authoritative_state', 'recursive_ip', 'authoritative_ip'])
else:
args.w = None
# Create main results directory if it doesn't exist
if not os.path.isdir('./results'):
os.mkdir('./results')
# Create results directory if it doesn't exist
if args.w is not None:
results_dir = './results/results_{}_{}_{}_{}/'.format(args.c, args.z, args.v, os.path.basename(args.w))
else:
results_dir = './results/results_{}_{}_{}_{}/'.format(args.c, args.z, args.v, "None")
if not os.path.isdir(results_dir):
os.mkdir(results_dir)
# Generating rankings, along with the count for each state
state_rankings = pairs.groupby(['recursive_state'])[['median']].agg([np.median, 'count'])
state_rankings.columns = state_rankings.columns.droplevel()
state_rankings = state_rankings[state_rankings['count'] >= 250]
state_rankings = state_rankings.sort_values('median')
state_rankings['rank'] = state_rankings['median'].rank(method='max', ascending=(column == 'rtt'))
state_to_state_p_values = pd.DataFrame()
state_to_state_p_values_i = pd.DataFrame()
state_to_state_h_values = pd.DataFrame()
# Generate kruskals values for each state pair
for i in range(0, state_rankings.shape[0]):
s1 = list(pd.DataFrame(state_rankings.iloc[[i]]).index)[0]
for j in range(0, state_rankings.shape[0]):
s2 = list(pd.DataFrame(state_rankings.iloc[[j]]).index)[0]
h, p = kruskal(pairs.loc[[s1]]['median'], pairs.loc[[s2]]['median'])
state_to_state_p_values.at[s1, s2] = p
state_to_state_p_values_i.at[s1, s2] = 1 - p
state_to_state_h_values.at[s1, s2] = h
# Generate clusters based on inverted kruskal's values
ac = AgglomerativeClustering(distance_threshold=.5, n_clusters=None, compute_full_tree=True).fit(state_to_state_p_values_i)
state_rankings['cluster'] = ac.labels_
# Save output
state_rankings.to_csv('{}/state_rankings.csv'.format(results_dir))
state_to_state_p_values.to_csv('{}/state_to_state_p_values.csv'.format(results_dir))
state_to_state_h_values.to_csv('{}/state_to_state_h_values.csv'.format(results_dir))
# Graph analysis
df = pd.read_csv('{}/state_to_state_p_values.csv'.format(results_dir))
df = df.rename(columns={'Unnamed: 0': 'states'})
df = df.set_index(['states'])
df[df > 0.05] = 1
df[df <= 0.05] = 0
# Create a network from the data
net = networkx.from_pandas_adjacency(df)
# Find cliques and convert to objects
raw_cliques = list(networkx.algorithms.community.greedy_modularity_communities(net))
cliques = [CliqueOfStates(i, l, [state_rankings.loc[state]['rank'] for state in l],
[state_rankings.loc[state]['median'] for state in l])
for i, l in enumerate(raw_cliques)]
cliques = sorted(cliques)
i = 0
colors = list(mcolors.TABLEAU_COLORS)
# Print cliques and generate intra-clique CDFs
for clique in cliques:
clique.set_id(i)
j = 0
for state in clique.states:
x = pairs.loc[pd.IndexSlice[state], :]['median']
color = colors[j % len(colors)]
plt.hist(x, bins=400, density=True, cumulative=True, label="{} ({:n})".format(state, clique.ranks[j]),
histtype='step', alpha=0.8, color=color, range=(0, 100))
j += 1
plt.legend(loc='upper left')
plt.savefig('{}/cluster_{}_cdf.png'.format(results_dir, i))
plt.show()
i += 1
print(clique)
print()
# Generate a CDF for each clique (all cliques)
for clique in cliques:
x = None
for state in clique.states:
if x is None:
x = pairs.loc[pd.IndexSlice[state], :]['median']
else:
x = pd.concat([x, pairs.loc[pd.IndexSlice[state], :]['median']])
color = colors[clique.cid % len(colors)]
plt.hist(x, bins=400, density=True, cumulative=True, label="{}".format('%s' % ', '.join(clique.states)),
histtype='step', alpha=0.8, color=color, range=(0, 100))
plt.axvline(x=np.median(x), color=color)
plt.legend(loc='upper left')
plt.savefig('{}/clusters_cdf.png'.format(results_dir))
plt.show()
@total_ordering
class CliqueOfStates:
def __init__(self, cid, states, ranks, values):
self.cid = cid
self.states = states
self.ranks = ranks
self.values = values
def set_id(self, cid):
self.cid = cid
def mean_value(self):
return np.mean(self.values)
def __str__(self):
state_list_string = ""
for state, rank, value in sorted(zip(self.states, self.ranks, self.values), key=lambda x: x[1]):
state_list_string += "\n\t{}\t{}\t{}".format(state, rank, value)
return "Clique {}: {}\nMean value: {}".format(self.cid, state_list_string, self.mean_value())
def __eq__(self, other):
return self.mean_value() == other.mean_value()
def __lt__(self, other):
return self.mean_value() < other.mean_value()
def get_or_create_pickle(column='rtt', z=2, cov_max=0.1):
if not os.path.isdir("./pickles"):
os.mkdir("pickles")
pickle_filename = './pickles/{}_{}_{}_pairs_pickle.zip'.format(column, z, cov_max)
if not os.path.isfile(pickle_filename):
print("Creating pickle...")
df = pd.read_csv('final_dns_dataset.csv', usecols=['rtt', 'authoritative_ip', 'recursive_ip', 'authoritative_state',
'recursive_state', 'rtt_normalized'])
# Generate z scores for individual measurements within server pairs
df['z_score'] = df.groupby(['recursive_state', 'authoritative_state',
'recursive_ip', 'authoritative_ip'])[column].apply(lambda x: (x - x.mean())/x.std())
# Filter out measurements that have z scores above the limit
df = df[abs(df['z_score']) <= z]
# Determine the coefficient of variation for a given series
def cov(x):
return np.std(x) / np.mean(x)
# Generate coefficient of variation for server pairs
pairs = df.groupby(['recursive_state', 'authoritative_state',
'recursive_ip', 'authoritative_ip'])[column].agg([np.median, cov])
# Filter out pairs with coefficient of variations too high - these are unreliable
pairs = pairs[pairs['cov'] < cov_max]
del pairs['cov']
pairs.to_pickle(pickle_filename)
return pickle_filename
if __name__ == "__main__":
main()