skywatch.blue
A tool for parsing traffic on the jetstream and applying a moderation workstream based on regexp based rules
Update facet detection to allow for exact duplicates
+40
-12
+16
-11
src/rules/facets/facets.ts
+16
-11
src/rules/facets/facets.ts
···
25
25
}
26
26
27
27
// Group mention facets by their byte position (byteStart:byteEnd)
28
-
// Only check mentions as duplicate tags/links are often bot bugs, not malicious
29
-
const positionMap = new Map<string, number>();
28
+
// Track unique DIDs per position - only flag if DIFFERENT DIDs at same position
29
+
// Same DID duplicated = bug, different DIDs = spam
30
+
const positionMap = new Map<string, Set<string>>();
30
31
31
32
for (const facet of facets) {
32
-
// Only count mentions for spam detection
33
-
const hasMention = facet.features.some(
33
+
// Only check mentions for spam detection
34
+
const mentionFeature = facet.features.find(
34
35
(feature) => feature.$type === "app.bsky.richtext.facet#mention"
35
36
);
36
37
37
-
if (hasMention) {
38
+
if (mentionFeature && "did" in mentionFeature) {
38
39
const key = `${facet.index.byteStart}:${facet.index.byteEnd}`;
39
-
positionMap.set(key, (positionMap.get(key) || 0) + 1);
40
+
if (!positionMap.has(key)) {
41
+
positionMap.set(key, new Set());
42
+
}
43
+
positionMap.get(key)!.add(mentionFeature.did as string);
40
44
}
41
45
}
42
46
43
-
// Check if any position has more than the threshold
44
-
for (const [position, count] of positionMap.entries()) {
45
-
if (count > FACET_SPAM_THRESHOLD) {
47
+
// Check if any position has more than the threshold unique DIDs
48
+
for (const [position, dids] of positionMap.entries()) {
49
+
const uniqueCount = dids.size;
50
+
if (uniqueCount > FACET_SPAM_THRESHOLD) {
46
51
logger.info(
47
52
{
48
53
process: "FACET_SPAM",
49
54
did,
50
55
atURI,
51
56
position,
52
-
count,
57
+
count: uniqueCount,
53
58
},
54
59
"Facet spam detected",
55
60
);
···
57
62
await createAccountLabel(
58
63
did,
59
64
FACET_SPAM_LABEL,
60
-
`${time}: ${FACET_SPAM_COMMENT} - ${count} facets at position ${position} in ${atURI}`,
65
+
`${time}: ${FACET_SPAM_COMMENT} - ${uniqueCount} unique mentions at position ${position} in ${atURI}`,
61
66
);
62
67
63
68
// Only label once per post even if multiple positions are suspicious
+24
-1
src/rules/facets/tests/facets.test.ts
+24
-1
src/rules/facets/tests/facets.test.ts
···
130
130
expect(createAccountLabel).not.toHaveBeenCalled();
131
131
expect(logger.info).not.toHaveBeenCalled();
132
132
});
133
+
134
+
it("should not label when same DID mentioned multiple times at same position (software bug)", async () => {
135
+
const facets: Facet[] = [
136
+
{
137
+
index: { byteStart: 0, byteEnd: 1 },
138
+
features: [{ $type: "app.bsky.richtext.facet#mention", did: "did:plc:user1" }],
139
+
},
140
+
{
141
+
index: { byteStart: 0, byteEnd: 1 },
142
+
features: [{ $type: "app.bsky.richtext.facet#mention", did: "did:plc:user1" }],
143
+
},
144
+
{
145
+
index: { byteStart: 0, byteEnd: 1 },
146
+
features: [{ $type: "app.bsky.richtext.facet#mention", did: "did:plc:user1" }],
147
+
},
148
+
];
149
+
150
+
await checkFacetSpam(TEST_DID, TEST_TIME, TEST_URI, facets);
151
+
152
+
// Should not trigger - only 1 unique DID
153
+
expect(createAccountLabel).not.toHaveBeenCalled();
154
+
expect(logger.info).not.toHaveBeenCalled();
155
+
});
133
156
});
134
157
135
158
describe("when spam is detected", () => {
···
161
184
expect(createAccountLabel).toHaveBeenCalledWith(
162
185
TEST_DID,
163
186
FACET_SPAM_LABEL,
164
-
`${TEST_TIME}: ${FACET_SPAM_COMMENT} - 2 facets at position 0:1 in ${TEST_URI}`
187
+
`${TEST_TIME}: ${FACET_SPAM_COMMENT} - 2 unique mentions at position 0:1 in ${TEST_URI}`
165
188
);
166
189
});
167
190