Gmail updated with new security measures, including ML-based early phishing detection

Google has announced a slew of new security features for Gmail, including early phishing detection based on machine learning, unintended external reply warnings, and other defenses against new threats.

To prevent phishing attacks, like this month’s Google Docs invite scam, Gmail is using a dedicated machine learning model that selectively delays emails (less than .05 percent of messages on average) to perform rigorous phishing analysis.

Integrated with Safe Browsing, a variety of techniques — like reputation and similarity analysis — will sift through for phishy and suspicious URLs. This works in conjunction with the previously announced click-time warning that redirects users before they visit phishing and malware links.

In a note to administrators, Google shares that as Safe Browsing has to test the results of links, emails can be delayed up to four minutes. As such, there is a setting to manually disable it for organizations that value speed.

Google notes that these machine learning models can adapt more quickly than a manual system ever could.

For G Suite customers, Gmail will now display unintended external reply warnings to users to help prevent data loss when responding to someone outside of a company’s domain. These warnings are smart and will avoid popping up for existing contacts and those that you regularly interact with.