Artificial Intelligence Provides New Path for Securing Connected Medical Devices

Posted June 4, 2018

Given the complexity inherent in today's world of increasingly connected medical devices, approaches using artificial intelligence (AI), machine learning, and deep learning will be needed to stay ahead of emerging cyberthreats, according to May Wang, PhD, chief technology officer and cofounder of Zingbox, Inc.

"We need to build trust in MDs," Wang said during her presentation on Monday at the AAMI 2018 Conference & Expo in Long Beach, CA. "But in this case, MD doesn't stand for 'medical doctors'; it stands for 'medical devices.'"

Current approaches to cybersecurity are associated with a number of technical and organizational challenges. For example, many end devices lack the capability for self-protection, such as the ability to download patches, according to Wang. That pushes the need for security to the network, which presents problems because firewalls tend to have a lot of vulnerabilities/weak points. Furthermore, the traditional approach of manually patching devices is no longer feasible given the sheer number of devices.

However, because IoT devices are built for specific purposes, their behavior can be known, Wang said. This means that they can be "taught" via machine learning.

The concept of deep learning, Wang explained, is nested in machine learning, which then is nested in AI. Deep learning can be used to build behavioral models (e.g., IoT device recognition, classification, and profiling).

Because AI is automatic, continuous, and proactive, forward-thinking cybersecurity approaches must leverage its capabilities, Wang advised. However, she said, this cannot happen without collaboration among various stakeholders (e.g., vendors, clinicians, healthcare technology management professionals, information technology professionals, regulators) and adequate investment into securing IoT devices.

By automating processes, AI has great potential to increase the amount of time health professionals have to focus on patient care.

"AI can do the boring tasks, allowing humans to do more creative, meaningful activities," she said.