Dev.to Security 🔐 Cybersecurity 👁 0 📖 2 min read

What a week of spam in crypto Telegram groups looks like, and the fake help questions behind it

If you run a Telegram group for a token or a crypto project, here is what a normal week of spam looked like in the groups we read, and which simple rules caught it. We read a week of messages from about 70 public Telegr

If you run a Telegram group for a token or a crypto project, here is what a normal week of spam looked like in the groups we read, and which simple rules caught it.

We read a week of messages from about 70 public Telegram groups, mostly crypto, that our reading account is a member of. Nothing was deleted, we only looked. Four things stood out.

1. The same message from many accounts

Close to one message in ten was the same text posted by several different accounts. About thirty texts, copied by well over a hundred accounts, filled a dozen groups.

What helps: a rule that notices when one text shows up from several new accounts in a short time, and removes all copies at once. Removing one copy by hand does not help, the next account posts it again.

2. Fake questions that set up fake admins

This one surprised us. In several groups, the questions that look like real members asking for help ("how do I claim my reward?", "where do I migrate my tokens?") were scripted. The same line came from different accounts, day after day, word for word. In one group, over two weeks, more than a hundred different reward and claim questions were each posted by two or more accounts.

The likely point is the reply: the fake question gives a fake helper a reason to answer with a link or a "message the admin" tip, and real members with the same question follow it.

What helps: treat the same help question from several accounts as a warning sign, not as a member in need. Pin one message that says real admins never DM first, and repeat it when these questions appear.

3. Shouted commands and keyword spam

Short shouted lines and messages full of trading or giveaway keywords were a small share; a plain word list was enough to flag them.

4. Where a simple filter gets it wrong

Two honest misses:

  • More than half of the messages that mentioned a recovery phrase were people warning others never to share it. A rule that deletes every mention of "seed phrase" deletes the messages of your most helpful members.
  • One of our link rules flagged normal links in Russian-language groups because it looked at the whole link, not only the site name. We changed the rule to look only at the site name.

So: test any filter on a week of your own group's messages before you let it delete anything.

Use something else when

  • Your group is small and quiet: Telegram's slow mode and a captcha on join are enough.
  • You only need to stop join raids: a captcha bot does that alone.
  • Your community lives on Discord: its built-in AutoMod covers most of this.

We run a community agent for token teams that does these checks, answers member questions from your docs, and tells the team when a human is needed. It is at fullyagentic.org.

📰 Read the original article on Dev.to Security

Originally published by Dev.to Security. Aggregated on AIWithGhost for educational purposes — full credit and traffic to the original publisher.