How Gmail uses machine learning to protect you from malicious emails.
A security blog posted by Google detailed how it effectively filters out various attacks on Gmail, and it is utilizing machine learning to effectively block more than 99.9 percent of threats that come through the email service.
Gmail already has various machine learning models in place that can effectively prevent phishing attempts, malware, and even spam. At the moment, more than 300 billion attachments are carefully scanned for harmful content and Google says that the malicious content can differ every single day.
Since these threats are evolving, Google must also add new features to its document scanners to keep up with the threat. This is why they are adding another layer for scanners to improve its detection capabilities. The new scanner was actually put in place in 2019, and this increased the threat detection rate by 150 percent.
Google goes into detail as to how these scanners work: it uses a “TensorFlow deep-learning model trained with TFX (TensorFlow Extended) and a custom document analyzer for each file type.” This basically lets the scanner parse the document to identify common attack patterns.
This new scanner runs in parallel with Gmail’s existing threat detection algorithms, which carefully filters out malicious content so that the only email you receive are the safe ones.
More details are available on Google’s Security Blog.