RSAC Fireside Chat: IRONSCALES utilizes LLM, superior intel to stay a step ahead of Deepfakes – Source: www.lastwatchdog.com

By Byron V. Acohido

There was a lot of buzz at RSAC 2024 about how GenAI and Large Language Models (LLM) are getting leveraged — by both attackers and defenders.

Related: Is your company moving too slow or too fast on GenAI?

One promising example of the latter comes from messaging security vendor IRONSCALES.

I had the chance to sit down with Eyal Benishti, IRONSCALES founder and CEO, to get a breakdown of how their new Generative Adversarial Network (GAN) technology utilizes a specialized LLM to reinforce anti-phishing mitigation services.

Benishti explained how GAN can very effectively mitigate Deepfaked messages, images, audio and video using a specially-tuned LLM to stay a step ahead of threat actors, even those who themselves are utilizing GenAI/LLM tools to enhance their Deepfakes.

Benishti told me about a remarkable GAN-powered phishing simulation test that took place recently with  highly-trained bank employees. GAN sent out personalized phishing ruses – and  deceived the employees who were previously impervious to template-based phishing tests.

One huge lesson gleaned is that the vendors who are integrating GenAI/LLM technology into their security tools have a huge advantage over threat actors: superior intelligence.

“We have access to public and non-public information, while the bad guys only have access to public information that anyone can get,” Benishti observes. “So we can really create something that is much more powerful than what they can create.”

When it comes to leveraging GenAI/LLM, it’s all about the prompting. For a full drill down, please give the accompanying podcast a listen.

Pulitzer Prize-winning business journalist Byron V. Acohido is dedicated to fostering public awareness about how to make the Internet as private and secure as it ought to be.

(LW provides consulting services to the vendors we cover.)

May 21st, 2024 | Fireside Chat | RSAC | Top Stories