Google AI scans billions of attachments in Gmail messages, identifies malicious documents

Google spoke about the work of the Gmail message attachment scanner, which is designed to ensure the safety of users by identifying malicious documents. Allegedly, such documents account for 58% of all malicious attacks aimed at Gmail users. Effective counteraction to malefactors became possible thanks to technologies of artificial intelligence.

Every week, one of Google’s main safeguards, the scanner processes over 300 billion attachments to block malicious content. At the same time, 63% of blocked malicious documents differ from those that were previously identified. To stay ahead of the curve, the scanner relies on deep learning.

According to Google, since the launch of the new scanner at the end of 2019, the daily reach of Office documents containing malicious scripts has increased by 10%. In addition to the training module, the scanner contains special analyzers for each type of file. They are responsible for parsing the document, identifying common attack patterns, extracting macros, de-obfuscating content, etc. So far, this technology is used only for scanning Office documents, but Google continues to actively develop it.