Credential-stealing emails are getting past artificial intelligence's "known good" email security controls by cloaking malicious payloads within seemingly benign emails. The tactic poses a significant threat to enterprise networks.
RETVec is a state-of-the-art text vectorizer which works directly on text inputs to create resilient classification models. Models trained with RETVec achieve better classification performance with fewer parameters and exhibit stronger resilience against adversarial attacks and typos, as reported in our paper.
Google is countering with RETVec (Resilient & Efficient Text Vectorizer). Open sourced by Google Research, this approach “helps models achieve state-of-the-art classification performance and drastically reduces computational cost,” while supporting “every language and all UTF-8 characters without the need for text preprocessing.” This makes it ideal for on-device, web, and other large-scale use cases:
