video1.0<iframe src="https://www.loom.com/embed/33f87e7836244b28ae054a346ce8ffff" frameborder="0" width="1920" height="1440" webkitallowfullscreen mozallowfullscreen allowfullscreen></iframe>14401920Loomhttps://www.loom.com14401920https://cdn.loom.com/sessions/thumbnails/33f87e7836244b28ae054a346ce8ffff-f092bc362295b7e3.gif785.82This Loom presents Smishi, a Serbian, Croatian, Bosnian, and Montenegrin SMS phishing detector, and demonstrates how it catches smishing patterns that are rising in Serbia. The speaker cites that smishing is up over 1,300 percent in Serbia in the last three years, and explains the system uses a fine-tuned Vertic model with 110 million parameters and reported performance around 96.96% accuracy and 96.3% F1. Examples cover Cyrillic and Latin scams impersonating Serbian institutions, suspicious domains such as .home and .help, typographical evasion with homoglyphs, urgency and deadline pressure, and sender-number inconsistencies. The Loom notes the app runs locally with no message leaving it and mentions 105 dataset test cases with 93.3% accuracy.