Skip to content
Locally-trained ML engine — no third-party AI

Detect Phishing Before It Detects You.

Analyze suspicious URLs, QR codes and emails using a locally developed cybersecurity intelligence engine — combining machine learning with independent heuristic analysis.

No credit card required · 2 free scans for guests

Drag & drop a QR image, or

Browse File
Analyze Email

Guests get 2 free scans. Register for more.

How PhishGuard Works

A transparent, multi-layer analysis pipeline — not a black box.

01

Submit

Paste a URL, upload a QR image, or paste email content.

02

Extract

Features are extracted — domain structure, entropy, QR payload, email headers.

03

Analyze

Our trained ML model and independent heuristic engine both score the input.

04

Report

A combined risk score, confidence level, and clear indicators are shown.

Three Ways We Protect You

Every input goes through the same rigorous risk engine.

URL Analysis

30+ engineered features — domain structure, entropy, suspicious keywords, TLD reputation and more — combined with independent heuristic rules.

QR Code Analysis

Decodes URL, text, Wi-Fi, and payment QR codes — including EMV/MMQR-style payment payloads — without guessing providers it can't confirm.

Email Analysis

Detects urgency language, credential requests, sender/display-name mismatches, and extracts every embedded URL for full analysis.

A Model Trained on Real Data — Not a Wrapper

Our detection engine is trained locally using scikit-learn, with domain-aware evaluation to prevent data leakage. We report real precision, recall and generalization — not inflated accuracy numbers.

  • Domain-grouped train/test splitting to avoid leakage
  • Calibrated probabilities, not raw model scores
  • Independent heuristic engine as a second opinion
model_version.json ACTIVE
algorithm
RandomForestClassifier
split_strategy
grouped_by_domain
calibration
isotonic
status
trained & evaluated

Full metrics published in the admin dashboard after training (Phase 12–13).

Built With Security First

No SSRF exposure

Server never blindly fetches user-submitted URLs against internal networks.

Privacy by design

Email bodies are hashed, not stored — only indicators are retained.

Role-based access

Every admin action is server-side authorized, not just UI-hidden.

Full audit trail

Logins, scans, and admin changes are logged for accountability.

Free vs Premium

Start free. Upgrade when you need more.

Free

For everyday personal checks.

  • ✓ Daily scan limit
  • ✓ URL, QR and Email analysis
  • ✓ Basic scan history
Get Started Free
RECOMMENDED

Premium

For power users and professionals.

  • ✓ Unlimited / high-volume scans
  • ✓ Advanced payment QR analysis
  • ✓ Detailed security reports
  • ✓ Priority feature access
Upgrade to Premium
3
Scan Types
30+
ML Features
0
Third-Party AI Used
100%
Local Analysis

Live scan statistics will populate here once real usage data is available.

Frequently Asked Questions

No. PhishGuard runs its own locally trained machine learning model and heuristic engine. No user data is sent to third-party AI APIs.
No. Some payment QR codes encode structured, tokenized data without a public brand name. In that case we report "provider could not be confidently identified" rather than guessing.
No. HTTPS is just one of over 30 signals we consider. A phishing site can also use HTTPS.
Guests get 2 free scans, tracked securely server-side. Registered free users get a daily limit; premium users get a much higher or unlimited limit.

Start Checking Before You Click.

Free to start. No credit card. Locally analyzed.