WhatsApp Is Testing Scam Alert: On-Device Warnings About Fraudsters

Translated by a model; the original is in Russian.

· Security
The optional feature uses a local ML model to flag messages from strangers that show signs of fraud. The contents of the conversation never leave the device.

cover.svg

WhatsApp has begun a limited beta test of a new optional feature called Scam Alert, which warns users about likely fraud attempts. For now the system is being road-tested together with researchers from the company's Bug Bounty program.

How it works

The key idea is that all processing happens locally. A machine learning model runs right on the device and analyzes incoming messages from people who aren't in your contact list. It's trained on scam conversations reported by users themselves, and it looks for familiar patterns, drawing on linguistic signals and a probabilistic classification of the conversation's structure.

If the model suspects deception, the user sees a warning in the chat offering to block the sender, report them, or keep chatting. The company stresses that message content never leaves the device for the sake of classification and isn't sent automatically to WhatsApp, to Meta, or to anyone else — the feature complements end-to-end encryption rather than undermining it.

If a warning turns out to be a false alarm, the chat can be marked as trusted: the notification then disappears and won't come up again. At the same time, the user can optionally share the last five messages received with WhatsApp to help improve the model's accuracy. Scam Alert can be turned off at any time.

Part of a bigger campaign against scammers

Scam Alert fits into a string of measures WhatsApp has been adding lately. In March, Meta announced that the messenger would begin warning about suspicious device-linking requests — a common account-hijacking trick in which the victim is tricked into scanning a malicious QR code or handing over a linking code.

Two months earlier came “strict account settings” — a lockdown-style enhanced protection mode for journalists, public figures, and other at-risk people. It was introduced after numerous cases of activists' and politicians' smartphones being infected with spyware, including NSO Group's Pegasus, via zero-click exploits that can compromise iOS or Android with no action at all on the owner's part.

WhatsApp is used by more than 3 billion people across over 180 countries, so even a narrowly targeted defense against scam messages reaches an enormous audience.


Source: Bleeping Computer

Comments

Sign in, to comment.

  • No comments yet.