Google tackles the amount of spam within Gmail even more. To do this, it integrates its machine learning framework TensorFlow with the mailapp. The new measures will complement existing Google algorithms against spam, and will further reduce the possibilities for spamming Gmail users.

In a blog post about the new features, product manager Neil Kumaran of Google’s Counter Abuse Technology department reports that the company can already block 99.99 percent of all spam, phishing attacks and malware that is trying to find its way to the users’ inbox. But Google is not easily satisfied and remains ambitious.

Defining spam mails

Google’s measures to deal with spam so far include machine learning algorithms and certain measures based on a fixed set of rules. Together they identify patterns in large datasets that people won’t see so easily. But thanks to many different factors, machine learning measures are relatively easy to find out what spam mails look like.

But, as Kumaran writes, just because some of the features of an e-mail are similar to those of what we would normally consider spam, that doesn’t mean that it’s really spam. Machine learning enables us to take all the signals together and make a decision based on them.

Even more ambition

However, the 99.99% that Google is currently filtering out is not enough for the company. That’s why with TensorFlow it developed a series of new ways to deal with spam. This allows the company to block an additional 100 million spam messages per day. Since we already block the majority of all spam messages in Gmail, it is quite an achievement to block millions more with precision, says Kumaran.

Above all, TensorFlow makes it possible to adapt the existing algorithms. This is how things get better and better at recognizing spam. It also helps to increase the speed with which these kinds of algorithms are built. They can therefore be trained more quickly and more easily, which makes tackling spam a lot more effective.

This news article was automatically translated from Dutch to give Techzine.eu a head start. All news articles after September 1, 2019 are written in native English and NOT translated. All our background stories are written in native English as well. For more information read our launch article.