After integrating AI-related features into all its productivity apps and Assistant, Google has released a new AI-based spam prevention algorithm for Gmail users. According to the Alphabet-owned company, the tool improves spam detection by 38 percent. Since the tool is developed using machine learning, it can also detect special characters. Here’s everything you need to know about Gmail’s new spam detection tool.
The New Spam Detection Tool Is Called RETVec
The new AI-powered spam detection tool is RETVec, which stands for Resilient & Efficient Text Vectorizer. Google is referring to it as its most significant security upgrade over the years.
Until now, Gmail was using the Unicode Transformation Format text detection for filtering certain words, which wasn’t effective against homoglyphs. However, RETVec can detect even subtle changes in fonts that scammers use and prevent users from interacting with the email.
RETVec Can Detect Spam In Over 100 Languages
Per Google, the new AI-based tool can detect spam in over 100 languages, including spam based on special characters. Further, RETVec won’t require as much processing power either. Google’s security blog confirms that even smartphones can run the algorithm to prevent spam.
It will also reduce dependency on Google’s servers by 78 percent, allowing the company to allocate the resources to a better place.
RETVec Improves Spam Detection And Requires Less Resources
Google has “battle-tested” RETVec extensively inside Google to evaluate how useful the tool is. In its testing, the company has found it to improve spam detection rate over baseline by 38 percent and reduce false positive rate by 194 percent.
Using the tool “reduced TpU usage of the model by 83%, making the RETVec deployment one of the largest defense upgrades in recent years.” Without text processing, the tool works out of the box on every language and all UTF-8 characters.