Watchlight
Watchlight is an iPhone app that checks texts for signs of scams and explains, in plain words, which warning signs it found. Every check runs on the phone. Watchlight has no server, no account and no analytics, so no message leaves the device.
I build Watchlight at Georgia State University as part of my research on how people fall for phishing, smishing (scam text messages) and voice scams. It is designed for older adults and for the relatives who help them with their phones.
What it does
- Sorts texts from unknown senders into Junk, Transactions or Promotions as they arrive, using Apple’s Message Filter. Sign-in codes are never sent to Junk.
- Checks anything you share with it: a text, link, QR code or screenshot. Each verdict lists the warning signs it found, how much each one counted, and what to do next.
- Brings in someone you trust: one tap sends the message and the verdict to a relative.
- Fits the person: read-aloud, large-text layouts, and English and Spanish.
- Leaves political speech alone: campaign texts are never labeled scams. You decide whether they are shown, sorted or sent to Junk.
How it is tested
Watchlight is small by design. Apple gives a message filter about 24 MB of memory and no network access, and Watchlight uses 8.9 MB. Every week, the frozen filter scores each new text from unknown senders, and only the counts are kept. So far it has not sent a legitimate text to Junk. However, it still catches too few new scams, so it sends a text to Junk only when its hand-checked rules are sure.
Status
Watchlight is in TestFlight beta. To join the beta or ask a question, email bampel@gsu.edu.