Because food delivery systems in the U.S. are so efficient, they are also potentially excellent bio-weapons delivery systems. Chip White, ISyE Schneider National Chair in Transportation and Logistics, and Alan Erera, ISyE Coca- Cola Professor, have teamed up to examine the relationship between the food supply chain and bioterrorism and how to mitigate the risk of a terrorist attack on the U.S. food supply chain.
“Because food supply chains are natural targets, one of the Department of Homeland Security’s (DHS) key priority areas is food systems,” said Erera. “Everyone eats. And the food delivery system is thought to be a compelling target for a terrorist attack that could lead to large numbers of sicknesses and fatalities, and could shatter people’s confidence in the food supply.”
Working with the DHS’s Food Protection and Defense Institute (FPDI) at the University of Minnesota, White and Erera have developed a model for countering an intelligent adversary who seeks to use a food supply chain as a means of delivering a biological or chemical toxin to the public-at-large, potentially leading to widespread illness and death.
“As a terrorist weapon, the idea of [contamination in the food system] is obviously one that can be quite powerful,” continued Erera. “You don’t want to worry about your food.
“This is something important that’s been on the minds of DHS for many years, and certainly since 9/11. Therefore, DHS prioritized food defense.”
White and Erera’s model is one of risk assessment and mitigation. It differs from what they call static models in that their model assumes the adversary — or would-be terrorist— is intelligent and adaptive.
“In our model, we have an adversary who is watching any defensive move to protect the food supply chain and is reacting and adapting accordingly,” explained White. “Traditional measures of risk don’t take such behavior into consideration.”
The risk model that the pair has developed assumes that because the adversary is intelligent and adaptive, he will change attack strategies or defensive postures over time, while the supply chain owner is doing likewise — what White described as a “cat-and-mouse game.”
The gaming aspect of the model involves both of the players — the adversaries and those who would thwart them — making the best decisions possible without knowing exactly what the other is doing. White and Erera apply an optimization framework to model how the players participate in the system over time and assess a related measure of risk.
It’s almost like a chess match: How do you choose the best way to act based on what your opponent does? However, in this case, you only have incomplete or partial knowledge of what your opponent is doing — “incomplete observations,” as White put it.
White and Erera have worked with FPDI for a decade and during that time have created different case studies to test their risk model. One such study involves feral swine and foot-and-mouth disease.
The wild swine, which could be infected with foot-and- mouth disease (intentionally or naturally), can then infect domestic herds — even when separated by fences — and a big outbreak of the illness could adversely affect the U.S. pork industry.
White said, “Feral herds have easy access to domestic herds and hence can serve as a way of infecting domestic herds with a variety of diseases, so there is quite a bit of concern about risks due to the intentional contamination of feral herds with foot-and-mouth disease.”
White pointed out that the model “is a framework, so it morphs with the application. Cargo theft is another application.” But even in this application, food security is still a concern: The product stolen most often is food. He continued, “Food can be stolen, contaminated, and sold on the gray market.”
White and Erera both emphasize that despite how it may sound, this risk assessment is not in any way simple. One of the innovations they’ve developed, according to Erera, “is embedding sequential stochastic optimization problems inside game theoretical decision models that have been around for a long time.
“Once optimal decision policies are determined, then you can think of it as a simulation,” Erera said. “Both of these players are playing this game optimally over time: One is trying to maximize the impact of attacks on the system; the other is trying both to balance the risk of an attack with maintaining productivity of the food supply chain. We simulate how the two opponents behave assuming that they’re going to behave near-optimally.
“And then we see how risky the system is under those settings, or decide whether the defender can take different actions, because they’re going to try some mitigation strategies. How much does that reduce these measures of risk? That’s basically how this model works.”
Stewart School of Industrial and Systems Engineering