When a message is confirmed to be spam, compute a fingerprint and persist it for that guild. If the same fingerprint is seen again, alert mods.
This would require changes to most existing cogs to centralize “this is a spam message” detection. Actual fingerprinting would be a new cog type.
Fingerprint can be calculated off:
message text
image sizes
image names
Think about how tolerant to make it. Perhaps calculate a few different fingerprints, e.g. with and without image file names. Or compute image fingerprints separately (filename + size).
When a message is confirmed to be spam, compute a fingerprint and persist it for that guild. If the same fingerprint is seen again, alert mods.
This would require changes to most existing cogs to centralize "this is a spam message" detection. Actual fingerprinting would be a new cog type.
Fingerprint can be calculated off:
- message text
- image sizes
- image names
Think about how tolerant to make it. Perhaps calculate a few different fingerprints, e.g. with and without image file names. Or compute image fingerprints separately (filename + size).
When a message is confirmed to be spam, compute a fingerprint and persist it for that guild. If the same fingerprint is seen again, alert mods.
This would require changes to most existing cogs to centralize “this is a spam message” detection. Actual fingerprinting would be a new cog type.
Fingerprint can be calculated off:
Think about how tolerant to make it. Perhaps calculate a few different fingerprints, e.g. with and without image file names. Or compute image fingerprints separately (filename + size).