Dr. Peter Beling is a professor in the Grado Department of Industrial and Systems Engineering at Virginia Tech and director of the Intelligent Systems Division of the Virginia Tech National Security Institute. He has contributed extensively to the development of methodologies and tools for the design and acquisition of cyber resilient military systems. His research interests include topics at the intersection of artificial intelligence and systems engineering. He earned a B.A. from the University of Virginia, an M.S. from the George Washington University, and a Ph.D. from the University of California at Berkeley.
Tell us about your current research – what excites you, what is challenging, and what impact are you motivated to achieve?
“My research explores the idea that we can engineer systems that benefit from new technologies while also managing the risks inherent to those technologies. For instance, many components and systems that used to be physical are now cyber-physical. Our cars now have dozens of computers and scores of sensors that are used to improve vehicle safety, improve fuel economy, and many other functions. These components are subject to cyber-attack that can lead to a catastrophic loss of system function. Similarly, the capability of artificial intelligence (AI) is growing exponentially, and most future systems will have key functions built around this technology. But AI is also subject to cyber-attack. Even without an adversarial action, the performance of AI systems can expect to fail over the lifecycle as system use or configuration drifts beyond the assumptions and data used in algorithm design or model training. My research focuses on methods for assuring overall system performance despite the fragilities of the new technologies at the heart of defense systems.”
As a researcher and member of the SERC Research Council, what has been your experience of collaborating with colleagues across the SERC network?
“The SERC network has provided me with the longest-lasting and deepest collaborations of my career. It’s a wonderful resource that is first in my mind when thinking about forming a proposal team or putting together a new professional event. The network helps us rise above institutional boundaries and function as a community of systems engineering researchers.”
Who most inspired you in your career, and what did you learn from them?
“Without question, the person who inspired me the most in my career is Dr. Barry Horowitz. I first met Barry when he joined the systems engineering faculty at the University of Virginia (UVA) after serving as CEO of MITRE Corporation and two startup companies. At the time, my department’s courses and research were narrowly focused on systems analysis (using primarily operations research) with some nod to systems thinking through case studies. Barry taught us that – even in an academic environment without access to large-scale systems – we could study the full range of systems engineering topics. His key lesson was that we needed to start working with technology in the form of Raspberry Pi, motes, and similar devices that cost tens of dollars, not tens of millions of dollars. That was only the start of Barry’s impact on my career, though. Barry was a pioneer of the Trusted Systems thrust within the SERC research roadmap and conceived the approach to resilience of cyber-physical systems that I am following in my SERC research today.”
Please give the SERC network recommendations for an interesting book, podcast, or article you’ve come across.
“I’d like to recommend a short video by Andrew Ng, a leading AI researcher. The video provides a sense of how vulnerable AI can be to adversarial attack. It’s important to know that similar loss of performance can happen through ordinary use of the AI. For instance, SERC researchers have shown that one can use machine learning to predict failures in machines based on sensor data (e.g., nose, vibration). But once the machine is repaired the accuracy of the machine learning model plummets to the point where it is unusable. That happens because the system has been changed by the maintenance action. The solution to this phenomenon lies more in improving system design or maintenance procedures than it does in improving the machine learning techniques.”