Scam Detector

Scam Detector

Detects scam images using a neural network (MobileNetV3).

(Scam images here means celebrity-impersonation scams made up of 4 images, or 2 images.)

Because it uses a neural network, it has a high detection rate and few false positives. However, since the training data is still limited, false positives may occur.

It also protects against threats beyond images.

・Dangerous URL detection: Checks URLs in messages against Google Safe Browsing and urlscan.io to catch phishing and fake sites.

・Malware file scanning: When executable files such as exe, scr, or zip are attached, they are scanned with MetaDefender Cloud.

・DM notice for hijacked accounts: When a malicious post is detected, the sender's account may have been hijacked, so the bot DMs them steps to reset their password and revoke authorized apps.

How to use

・Just add it to your server and it works with no extra setup. Detected messages are deleted automatically. (To timeout, kick, or ban members, the bot's role must be placed higher than regular members' roles.)

・On servers with a verification system, the bot generally needs the verified role to work.

・The bot cannot detect scams in channels it cannot access.

・You can change settings with the /settings command. (Admins only)

/settings action: Set the action on detection (delete only / timeout / kick / ban)

/settings log-channel: Set the channel for detection logs (omit to disable)

/settings language: Set the display language (follow server locale, or choose from 13 languages)

/settings show: Show the current settings

・If a scam image was missed, admins can report it by right-clicking the message → Apps → "Report Scam".