GitHub Pages edition. Scoring, character attacks, normalization, and bounded email preview run entirely in this tab. Run the full local API for RFC-complete triage, semantic attacks, and strict LureProof evidence.

Defensive AI engineering / live workbench

Break the detector before they do.

Score a suspicious message, mutate it like an attacker, then verify whether your defense earns the catch back.

01 / Input

Message under review

Processed locally

Message text

Detection engine

02 / Readout

Detection signal

Awaiting analysis

Run the message through a detector
to reveal its fraud signal.

—%

BENIGN / 0THRESHOLD / 50FRAUD / 100

Detector—

Threshold source—

Contributing terms

Real-world workflow / RFC 5322

Triage a suspicious email file

Choose a saved .eml. LureScope reads its message text, headers, link names, and attachment names locally. It never visits links or opens attachment contents.

Choose .eml file

—

——

—

No message content or addresses are written to the proof.

03 / Stress test

Run the adversarial sequence

No external calls for character attacks

Attack technique

Countermeasure

01 / Original—

02 / Attacked—

03 / Defended—