Fraudster Alert: AI Fueling New Gmail Upgrade

Google is claiming its new Gmail move will offer users “the best protection” ever.

Email fraud, especially phishing scams, is one of the most insidious and damaging forms of digital crime.

One big reason why? Email fraud fees are almost entirely on human error, even though most users try to cover their tracks, reinforce their password security, and otherwise lean into secure email software user recommendations.

According to industry statistics . . .

• A projected 3.4 billion emails daily are transmitted by cyber fraudsters. That amounts to over one trillion fraudulent emails annually.

• Email impersonators are responsible for approximately 1.2% of all global emails on an annual basis.

• Phishing accounts for 36% of data breaches.

• 84% of organizations were targeted by phishing fraudsters last year. That’s up 15% from 2021.

Now, Google is introducing a new AI-based technology that takes dead aim at email fraudsters.

The new AI technology tool – called Resilient & Efficient Text Vectorizer (RETVec) – helps email models achieve “state-of-the-art classification performance and drastically reduces the computational cost,” says Elie Bursztein, Google cybersecurity and AI research director in a company statement. “Today, we’re sharing how RETVec has been used to help protect Gmail inboxes.”

That strategy focuses on text classifiers more robust and efficient and Google has developed a novel, multilingual text vectorizer to get the job done.

“Systems such as Gmail, YouTube, and Google Play rely on text classification models to identify harmful content including phishing attacks, inappropriate comments, and scams,” says Marina Zhang, a Google software engineer. “These types of texts are harder for machine learning models to classify because bad actors rely on adversarial text manipulations to actively attempt to evade the classifiers. For example, they will use homoglyphs, invisible characters, and keyword stuffing to bypass defenses.”

That’s where RETVec’s machine learning software, rolling out this week, changes the equation.

“Highly Effective”

Over the past year, Bursztein says Google has” battle-tested RETVec extensively” to evaluate its usefulness. In doing so, Google engineers have found the technology to be “highly effective for email security and anti-abuse applications.”

“In particular, replacing the Gmail spam classifier’s previous text vectorizer with RETVec allowed us to improve the spam detection rate over the baseline by 38% and reduce the false positive rate by 19.4%,” Zhang notes.

Google says that leveraging RETVec reduced the TPU usage of the model by 83%, making the RETVec deployment “one of the largest defense upgrades in recent years.”

“RETVec achieves these improvements by sporting a very lightweight word embedding model, allowing us to reduce the Transformer model’s size at equal or better performance, and having the ability to split the computation between the host and TPU in a network and memory efficient manner,” Bursztein adds.

Consequently, if you notice Google taking a more aggressive stance with Gmail scam alerts, you’ll know why.

Fraudsters, too, should take notice. Now they’re in the crosshairs instead of the other way around.



Recent Posts

Leave a Reply

Your email address will not be published. Required fields are marked *