Explainability
PHISHLENS explains why.
Your contribution is not just detection — it’s interpretable evidence. This supports user trust, decision-making, and reduces “black-box” risk.
Evidence tokensReason categoriesHighlighted textConfidence
1) Evidence tokens
The model returns influential words/phrases contributing to the prediction.
2) Reason categories
Signals are grouped into user-friendly categories like urgency, credentials, links, and threats.
3) Highlighted body
Suspicious phrases are highlighted directly in the email text to make the decision transparent.