Google has added some new security features to Gmail, and one of them reportedly improves phishing detection by using a specialized machine learning model.
According to the company, 50 to 70 percent of incoming messages in Gmail are spam, but machine learning techniques help Google beat spam and phishing messages with an accuracy of up to 99.9 percent.
The latest machine learning model improves the process by delaying selected messages (less than 0.05% of messages on average) for further phishing analysis.
Google says the new detection models will also help create new warnings by delaying URL access time, showing users a warning prompt when they've clicked on a suspicious link. As new patterns are identified, the new protection models will adapt and improve over time.
Additionally, Google added external reply warnings to company email to prevent users from inadvertently sending protected data to someone outside their company.
Gmail will not stop the sending but will try to delay it a bit with warnings, according to Google, as machine intelligence will notify Gmail if the recipient is an existing contact or someone with whom the sender interacts regularly.
Google also added new built-in defenses against ransomware and polymorphic malware, blocking millions of add-ons that are distributed as email attachments.
Threats are identified by combining flags from spam, malware, and ransomware with attachment heuristics (emails that may be threats based on the flags) and the digital signatures of senders (who are already marked as malware).
