Eric Chien: Machine Learning-Based Tools Could Help Agencies Disrupt Attackers

Eric Chien/LinkedIn

Eric Chien, a fellow on the Threat Hunter Team at Broadcom, said government agencies looking to counter attackers should create networks that are secure by default and use tools that leverage artificial intelligence and machine learning.

Chien noted that achieving a network that is secure by default means reducing the attack surface and adopting zero trust principles, microsegmentation and multifactor authentication.

Security Model Tailored to an Organization’s Environment

In an article published on Carahsoft.com, Chien cited technologies like Broadcom’s Symantec Adaptive Security and how such tools leverage machine learning to prevent cyberattacks without harming productivity.

He said the Symantec Adaptive Security platform “flips the model of security from one-size-fits-all to an approach that is dynamically tailored to the organization’s environment and uses machine learning to block everything except the normal business processes that a user or device may need to do,” he wrote.

Taking Remediation to the Next Level

In this piece, Chien said Broadcom has developed a tool, called Incident Prediction, that could help organizations prevent the need to shut down entire networks while IT administrators focus on tracking down cyberattackers.

He noted that Incident Prediction, which is based on large language models, could help organizations predict and block a threat actor’s next move.

“By training our own analogous LLM on half a million attack chains collected in just the previous six months, we developed a tool that can predict the next four or five steps an attacker will take with 100% confidence,” Chien wrote.

“Incident Prediction can also identify what an attacker might do in response to those defensive actions so the security team can block those as well,” he added.