In his own words, Google filters out now well 99.9% of spam messages sent to Gmail mailboxes. The false positive rate – the number of cases where a regular mail as spam has been mistakenly classified – was now less than 0.05 percent. These values resulted from the use of an artificial neural network, with the Google billions of incoming news searches to weed out unwanted mails and phishing attacks such as product manager Sri Harsha Somanchi writes in a blog post .
Would do everything in his power Google to combat spam, Gmail would probably not reasonably usable. According to the Russian security vendor Kaspersky, were 59.2 percent of all email spam messages filtered out in the first quarter by 2015. The sender be changed even extra to newly released top-level domains as .work or .science to bypass spam filters and deliver advertising or malware.
Since the early days of Gmail , Google relies on machine learning. While it relies on its 900 million people who help improve the SpamFilter by manually logging messages as spam or use of the mark “Not spam” button. The latter is sometimes still necessary to Somanchi according to. So, users may need to crawl their complete spam folder, to find an important E-Mail, which was – accidentally classified as spam, for example, a monthly statement of their bank.
To address this problem, Google leads a new system called Gmail Postmaster tools a. Reputable companies who send emails en masse, can evaluate data about delivery errors, spam reports, and their reputation with him. In this way, a problem can identify and make sure that Gmail will forward the messages to the right place. For users, this means Google according to “no further rummaging in the trash”.
In his blog post, Somanchi explains in broad, like Google’s artificial neural network, which also comes in products such as the search and Google now used, helps to combat “especially sneaky spam”. “Thanks to new machine-learning hints Gmail can find out now whether a message actually comes from the supposed sender and curb fraudulent E-Mails”, so the product manager. Machine learning uses Google in its spam filters also help to recognize the different tastes of different Gmail users. While an approximately weekly newsletter is located in his mailbox, another may not be available from such messages.