Google has created an AI-driven spam detection system to make Gmail safer


Google has created an AI-driven spam detection system to make Gmail safer
Google has developed an AI-powered spam detection system that helps detect "adversarial text manipulations", such as emails with special characters, emojis, typos and other characters that easily bypass Gmail's protections. Google's update, billed as "one of the largest defense upgrades in recent years", introduces RETVec (Resilience and Effective Text Vectorization) as its new text classification system.
"To help make text classifiers more robust and efficient, we’ve developed a novel, multilingual text vectorizer called RETVec that helps models achieve state-of-the-art classification performance and drastically reduces the computational cost", the company said. Systems such as Gmail, YouTube and Google Play rely on text classification models to identify harmful content including phishing attacks, inappropriate comments, and scams.
Identifying these kinds of texts poses a challenge for machine learning models, as malicious entities employ adversarial text manipulations to actively evade classifiers. The use of tactics like homoglyphs, invisible characters, and keyword stuffing is common among these bad actors to bypass defensive measures, as highlighted by the tech giant. RETVec, with its innovative architecture, stands out by seamlessly functioning across all languages and character sets without necessitating prior text preprocessing. This characteristic positions RETVec as the optimal choice for on-device, web, and large-scale text classification deployments, as it is adept at handling diverse linguistic nuances and ensuring robust defense against adversarial tactics.
"Models trained with RETVec exhibit faster inference speed due to its compact representation. Having smaller models reduces computational costs and decreases latency, which is critical for large-scale applications and on-device models", the tech giant said. RETVec is a novel open-source text vectorizer that allows people to build more resilient and efficient server-side and on-device text classifiers. The Gmail spam filter uses it to help protect Gmail inboxes against malicious emails, said Google.
Source: IANS