Created by Sharvesh

Waking the threat engine…0%

PHISHGUARD

AI · EMAIL SECURITY · LIVE 3D

PHISHGUARD

Machine-learning phishing detection that shows its work — scoring every email on text, links and sender trust.

Scan an email↓
MODEL—
ACCURACY—
EMAILS TRAINED—
FEATURES—
01 / INPUTCTRL+ENTER scan · CTRL+V paste image
Drop a screenshot here or click to browse · or press Ctrl+V to paste one
STANDBY
AWAITING
0.0% phishing risk
02 / EVIDENCE—

No email scanned yet.
Pick a sample or paste one to see why it's flagged.

03

Model telemetry

Measured on a held-out test split the model never saw during training or model selection.

CONFUSION MATRIXrows = actual

Heads-up: the bundled dataset is synthetic, so scores are near-perfect by design. Swap in a real corpus (see README) for realistic numbers.

04

How it works

01

Parse

Extracts links, sender domain, brand mentions and urgency language from the raw email.

02

Featurise

Combines TF-IDF text features with 21 handcrafted signals — look-alike domains, IP links, shorteners, bad TLDs.

03

Classify

A cross-validated Logistic Regression / Random Forest picks the phishing probability.

04

Explain

Every verdict ships with the exact red (and green) flags behind it — no black box.