Google recently developed and unveiled RETVec, a resilient and efficient text vectorizer to defend Gmail users against malicious emails and spam. RETVec is a next-gen text vectorizer with built-in adversarial resilience that is able to handle character-level manipulations like insertion, deletion, typos, homoglyphs, and LEET substitution. It supports 100+ languages without lookup tables or fixed vocabulary and can be easily employed in TensorFlow models for string vectorization. RETVec improves spam detection by 38% with a 0.80% false positive rate, reducing latency by 30%. It is competitive for real-world tasks and eliminates the need for a vocabulary layer in smaller language models. However, optimizing RETVec for multilingual skills, robustness, and smaller models in large language models poses challenges, as does its use in generative models. Future work will address these limitations and explore RETVec's potential as a word embedding tool. #cybersecurity #RETVec

https://cybersecuritynews.com/google-revealed-retvec-to-defend-malicious-emails-spam-for-gmail-users/