Google has been using artificial intelligence to teach cars to drive themselves, to recognize human speech and to mimic the way we see. But is artificial intelligence really smart enough to help us get rid of spam?
In a post on the Official Gmail blog, the company said it already has a low spam rate of less than 0.1 percent, in part thanks to the way users train its system by labeling messages as spam or not spam. More recently, however, Google started using an artificial neural network — a kind of software that mimics the way the human brain learns — to filter out certain kinds of messages. This could include phishing emails and other communications that try to dupe victims into clicking through to malware or sites that could take over their machine and harvest their data.
As ZDNet and others pointed out, Google has yet to offer any hard numbers on the volume of phishing emails it has managed to block. Unless users reported failed attempts using such spam, it might be difficult to do so. However, even noticing trends in emails that try to impersonate an individual or an organization such as a bank would be a huge boon for many consumers targeted by cybercriminals.
Of course, even without an artificial neural network, there is data that could help Google and other companies better protect users from the most harmful forms of spam. A story on eWEEK cited a study that showed senders of unsolicited emails largely come from the U.S., Europe and China. It would be interesting to know to what extent Gmail’s machine learning features are integrating third-party research to boost security even further.
Another way to fight back against attackers using email is to improve the way legitimate organizations send messages to customers and prospects. This comes in the form of Gmail Postmaster Tools, which VentureBeat described as a way for enterprises to analyze the results of their email campaigns and ensure what they send isn’t treated as spam.
If all this works, expect more pressure on other organizations to beef up the way they allow messages to move across their platforms. In fact, Wired noted how Facebook has been using machine learning for similar purposes, while Twitter and even Chinese search engine Baidu are following suit. Malicious spammers may finally start getting the message.