Email address scoring ML model

This page enables you to test how suspicious/fraudulent an email looks. Behind the hood, it uses a supervised ML model that leverages only signals/features related to the email string, such as:

• Whether or not the email domain is disposable;
• The popularity of the email provider (hotmail, gmail, etc);
• The appearance of the username (the part before the @).

The ML model was trained on freely available data and DOESN'T leverage ANY kind of private user/email history. Thus, you should be careful how you interpret this score.

You can also access this data through this API.

How to interpret the risk score?

A risk score close to 1 indicates a potentially suspicious/fraudulent email, while a risk score close to 0 indicates a more legitimate email. As indicated previously, the ML model only analyzes the appearance of the email address only with information related to the email domain/provider. Thus, you should take into account that the score DOESN'T reflect any information about the actual user history or intent.