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
|
"""Logic to perform spam/ham classification"""
import sys
from typing import Any, List
from google.protobuf.json_format import MessageToDict
from vyper import v
from api.v1.spamcheck_pb2 import SpamVerdict
from app import event, logger, queue, data_store
from server.interceptors import SpamCheckContext
log = logger.logger
classifiers = v.get_string("ml_classifiers")
if classifiers:
sys.path.append(classifiers)
# pylint: disable=too-few-public-methods
class Spammable:
"""Base class for spammable types."""
allow_list = v.get("filter.allow_list")
deny_list = v.get("filter.deny_list")
allowed_domains = set(v.get("filter.allowed_domains"))
# Currently maximum allowed value is conditional allow to limit false positives.
max_verdict = SpamVerdict.CONDITIONAL_ALLOW
_inference_scores = {
0.9: SpamVerdict.BLOCK,
0.5: SpamVerdict.DISALLOW,
0.4: SpamVerdict.CONDITIONAL_ALLOW,
0.0: SpamVerdict.ALLOW,
}
_verdict_rankings = {
SpamVerdict.ALLOW: 1,
SpamVerdict.CONDITIONAL_ALLOW: 2,
SpamVerdict.DISALLOW: 3,
SpamVerdict.BLOCK: 4,
}
_verdict_mapping = {
"ALLOW": SpamVerdict.ALLOW,
"CONDITIONAL_ALLOW": SpamVerdict.CONDITIONAL_ALLOW,
"DISALLOW": SpamVerdict.DISALLOW,
"BLOCK": SpamVerdict.BLOCK,
}
def __init__(
self, spammable: Any, context: SpamCheckContext, classifier: None
) -> None:
self.context = context
self.spammable = spammable
self.classifier = classifier
@classmethod
def set_max_verdict(cls):
"""Set the maximum verdict for the spammable class."""
key = f"max_{cls.__name__}_verdict".lower()
verdict_value = v.get_string(key).upper()
try:
cls.max_verdict = cls._verdict_mapping[verdict_value]
except KeyError:
valid_args = ', '.join(cls._verdict_mapping.keys())
log.fatal(f"Max verdict must be in [{valid_args}]. Got: {verdict_value}")
@property
def spammable(self) -> Any:
"""spam.Spammable: The spammable to analyze for spam"""
return self._spammable
@spammable.setter
def spammable(self, spammable: Any):
self._spammable = spammable
self._email_allowed = self.email_allowed(spammable.user.emails)
if spammable.project:
self._project_allowed = self.project_allowed(spammable.project.project_id)
else:
self._project_allowed = True
def verdict(self) -> SpamVerdict:
"""Analyze the spammable and determine if spam.
Returns:
SpamVerdict
"""
# If the project is not allowed then this may be an indication that the model
# does not generalize well to the spammables in that project. In this case we will
# circumvent evaluating the spammable.
if not self._project_allowed:
return self._verdict(SpamVerdict.NOOP, 0.0, "project not allowed", False)
if not self.classifier:
return self._verdict(SpamVerdict.NOOP, 0.0, "classifier not loaded", False)
spammable_dict = self.to_dict()
confidence = self.classifier.score(spammable_dict)
data_store.save(self.type(), spammable_dict, confidence)
if self._email_allowed:
return self._verdict(SpamVerdict.ALLOW, confidence, "email allowed", True)
verdict = self.calculate_verdict(confidence)
return self._verdict(verdict, confidence, "ml inference score", True)
def calculate_verdict(self, confidence: float) -> SpamVerdict:
"""Convert an ML confidence value to a spam verdict.
Args:
confidence (float): The ML confidence value
Returns:
SpamVerdict
"""
for threshold, vdict in self._inference_scores.items():
if confidence >= threshold:
return self._maximum_verdict(vdict)
return SpamVerdict.NOOP
def _verdict(
self, verdict: int, confidence: float, reason: str, evaluated: bool
) -> SpamVerdict:
fields = {
"correlation_id": str(self.context.correlation_id),
"metric": "spamcheck_verdicts",
"spammable_type": self.type(),
"email_allowlisted": self._email_allowed,
"project_allowed": self._project_allowed,
"project_path": self._spammable.project.project_path,
"project_id": self._spammable.project.project_id,
"user_name": self._spammable.user.username,
"user_in_project": self._spammable.user_in_project,
"verdict": SpamVerdict.Verdict.Name(verdict),
"reason": reason,
"confidence": confidence,
"evaluated": evaluated,
}
log.info("Verdict calculated", extra=fields)
if verdict not in (SpamVerdict.ALLOW, SpamVerdict.NOOP):
evnt = event.Event(event.VERDICT, fields)
queue.publish(evnt)
return SpamVerdict(
verdict=verdict, score=confidence, reason=reason, evaluated=evaluated
)
def project_allowed(self, project_id: int) -> bool:
"""Determine if a project should be tested for spam.
Args:
project_id (int): The GitLab project ID
Returns:
bool
"""
if len(self.allow_list) != 0:
if self.allow_list.get(project_id) is not None:
return True
return False
if len(self.deny_list) != 0:
if self.deny_list.get(project_id) is not None:
return False
return True
return True
def email_allowed(self, emails: List) -> bool:
"""Determine if a user email should be exempt from spam checking.
Args:
emails (list): A list of Emails represented by protobuf objects
Returns:
bool
"""
for email in emails:
if not "@" in email.email:
continue
domain = email.email.split("@")[-1]
if email.verified and domain in self.allowed_domains:
return True
return False
def type(self) -> str:
"""Get the string representation of the spammable type."""
return type(self).__name__.lower()
def to_dict(self) -> dict:
"""Return the dictionary representation of the spammable."""
spammable_dict = MessageToDict(self._spammable)
spammable_dict["correlation_id"] = str(self.context.correlation_id)
return spammable_dict
def _maximum_verdict(self, verdict: SpamVerdict) -> SpamVerdict:
max_verdict = self._verdict_rankings[self.max_verdict]
current_verdict = self._verdict_rankings[verdict]
if max_verdict < current_verdict:
return self.max_verdict
return verdict
|
