Addressing cybersecurity concerns in IoT

The Internet of Things is coming of age.

The explosion of connected devices—and their rapid insurgence into our homes and workplaces—has so far outpaced investment in keeping them secure.

But that’s starting to change, as buyers and sellers of “smart” products are less willing to throw caution to the wind when it comes to cybersecurity, according to Dr. Rick Blum, the Robert W. Wieseman Professor of Electrical Engineering at Lehigh University. 

Whereas systems like the energy grid are regulated and “are going to be protected,” Blum explains, “the Internet of Things systems—when you’re talking about self-driving cars or sensors in your body—these are the systems we better worry about.”

Blum, who is affiliated with the university's Institute for Cyber Physical Infrastructure and Energy (I-CPIE), a hub for interdisciplinary research, says it’s critical that developers address security concerns during the system or product design process and don’t cut corners for the sake of cost. 

For example, in the case of self-driving cars, he says, manufacturers use a combination of radar data and video camera data to prevent crashes with other vehicles or pedestrians. However, radar systems are known to be vulnerable to remote attacks.

One way to address these vulnerabilities, Blum says, is to protect the system on multiple fronts.

“We want to have many, many layers of security. And if some of these layers are breached, we want there to be some of these other layers available to provide protection,” he says.

The layers, which could include firewalls, encryption, and other types of protection, will vary depending on the commercial application, and developing them will require field-specific expertise.

Blum’s team is developing the theory of cybersecurity for cyber physical and Internet of Things systems for future textbooks. They are also developing novel machine learning-based algorithms to protect these systems. These algorithms employ all available information: including mathematical models, historical data, forecasts, and sensor measurements.