Google bolsters Gmail spam detection with artificial neural networks

By Shawn Knight
Jul 10, 2015
Post New Reply
  1. Messaging apps may be getting all of the media attention these days but the reality is, e-mail continues to be an integral facet of the Internet. Another (unfortunate) reality is that spam continues to plague our inboxes, serving as a minor distraction to some and a source of pure frustration for others.

    Gmail is often considered one of the best e-mail clients as it relates to anti-spam measures. Being ahead of the curve can either mean you rest comfortably on your laurels or continue pushing ahead and advancing your position. Fortunately for Gmail users, Google subscribes to the latter philosophy.

    In a recent blog post, Google product manager Sri Harsha Somanchi said 0.1 percent of messages in the average Gmail inbox are spam and less than 0.05 percent of legitimate e-mails find their way to the spam folder. Even still, there’s room for improvement.

    Somanchi noted that Gmail is now using an artificial neural network – like what it uses to power Google Search and Google Now – to help detect and block especially sneaky spam. Gmail’s spam filter works on an individual basis, learning the types of messages you do and don’t like over time.

    Your neighbor, for example, may love weekly newsletters sent to his inbox while you can’t stand them. The spam filter can now pick up and act on such preferences.

    What’s more, the spam filter is better than ever at detecting phishing attempts. By using new machine learning signals, Google said it can now figure out whether or not a message actually came from its perceived sender to help keep bogus messages out of your life.

    Google’s improvements aren’t limited to those on the receiving end of messages. The company has launched a new product called Gmail Postmaster Tools that’s designed to help qualified high-volume senders like airlines and banks analyze their messages, diagnose issues, study best practices and above all else, make sure legitimate messages make it to their destinations.

    Permalink to story.


Similar Topics

Add New Comment

You need to be a member to leave a comment. Join thousands of tech enthusiasts and participate.
TechSpot Account You may also...