Autospam
+The automated spam detection system
+Active Autospam
+ {{ formatCount(config.open) }} +Closed Autospam
+ {{ formatCount(config.closed) }} ++
Autospam Service Operational
+Stats
+Autospam Detections
+About Autospam
+ +To detect and mitigate spam, we built Autospam, an internal tool that uses NLP and other behavioural metrics to classify potential spam posts.
+ ++ +
Standard Detection
+ +Standard or "Classic" detection works by evaluating several "signals" from the post and it's associated account.
+ +Some of the following "signals" may trigger a positive detection from public posts:
+ +-
+
- Account is less than 6 months old +
- Account has less than 100 followers +
- Post contains one or more of: https:// http:// hxxps:// hxxp:// www. .com .net .org +
If you've marked atleast one positive detection from an account as Not spam, any future posts they create will skip detection.
+ ++ +
Advanced Detection
+ +Advanced Detection works by using a statistical method that combines prior knowledge and observed data to estimate an average value. It assigns weights to both the prior knowledge and the observed data, allowing for a more informed and reliable estimation that adapts to new information.
+ +When you train Spam or Not Spam data, the caption is broken up into words (tokens) and are counted (weights) and then stored in the appropriate category (Spam or Not Spam).
+ +The training data is then used to classify spam on future posts (captions) by calculating each token and associated weights and comparing it to known categories (Spam or Not Spam).
+ ++ In order for Autospam to be effective, you need to train it by classifying data as spam or not-spam. +
+ ++ We recommend atleast 200 classifications for both spam and not-spam, it is important to train Autospam on both so you get more accurate results. +
++ +
+ +Use existing posts marked as spam to train Autospam
+ + ++ +
+ +Use posts from trusted users to train non-spam posts
+ + +| ID | +Type | +Reported Account | +Created | +View Report | +
|---|---|---|---|---|
| + {{ report.id}} + | +
+ Autospam Post + |
+
+
+
+
+
+
+
+ @{{report.status.account.username}} +
+ {{report.status.account.followers_count}} Followers
+ ·
+ Joined {{ timeAgo(report.status.account.created_at) }}
+
+ |
+ {{ timeAgo(report.created_at) }} | +View | +
+ Tokens are used to split paragraphs and sentences into smaller units that can be more easily assigned meaning. +
+| ID | +Token | +Category | +Weight | +Created | +Edit | +
|---|---|---|---|---|---|
| + {{ token.id}} + | +
+ {{ token.token }} + |
+
+ {{ token.category }} + |
+
+ {{ token.weight }} + |
+ {{ timeAgo(token.created_at) }} | ++ Edit + | +
+ +
+No custom tokens found!
++ You can import and export Spam training data +
+ ++ We recommend exercising caution when importing training data from untrusted parties! +
++ +
+ +Make sure the file you are importing is a valid training data export!
+ + ++ +
+ +Only share training data with people you trust. It can be used by spammers to bypass detection!
+ + +Select trusted accounts to train non-spam posts against!
+Token
+Weight
+Category
+Note
+Active
+Token
+Weight
+Category
+Note
+Active
+