Media
3 items
Description
Spots impersonation and phishing in your inbox, explains why, and stops you before you type a password or code into a fake page. Impersonation Guard protects you, not a company. It reads the message you have open, tells you in one sentence whether it is safe and exactly why, and interrupts at the moment of harm: before you type a password or a one-time code into a page that only looks like the real site. WHAT YOU SEE • In Gmail and Outlook on the web, a banner above every open message: Looks OK, Be careful, Likely a scam, or Scam. Under it, one plain sentence per finding and the exact thing to look at: the sender's name beside the real address, a link's domain beside the brand it imitates, the thing the message asks you to do. • On any page, if a form asks for a password or a one-time code and the page's address imitates a site you use, or you got there from a message the guard flagged, a full-screen stop blocks the submission until you decide. • In the toolbar popup, paste a text message, a WhatsApp or Discord message, or drop a screenshot. The text is read on your computer and checked the same way. HOW IT DECIDES Three layers, combined into one verdict. Lookalike names: Unicode confusable characters, mixed scripts, punycode and one-or-two-letter typos against the names you protect. Sender score: a small explainable model over 23 named signals, such as a brand name on an address the brand does not own, or an official-sounding role from a free webmail account. Message intent: what the message asks for, from a rule engine or, if you turn it on, a language model running locally through Ollama. Which names are checked is decided per message: your own list in Settings, the domains of your own accounts, and any well-known brand the message mentions or links to. PRIVACY Everything runs inside the extension on your computer. No message, screenshot or address leaves it. The extension makes no network requests of its own. The only exception is one you turn on yourself: the optional local AI model, which sends text to the Ollama server on your own machine. No analytics, no accounts, no telemetry. The code is open source under the MIT license. WHAT IT DOES NOT DO It does not replace your mail provider's spam filter or your browser's Safe Browsing; it sits after them and explains what they do not. It does not know every brand; add your school, employer and bank in Settings. A rule engine can miss lures written in an unfamiliar register; the optional local model catches more. Source and documentation: https://github.com/Sonchez1/impersonation-guard