This talk considers information fusion problems embedded in national critical infrastructure. We discuss three current research problems: 1. Detection of radiation sources - Reliable detection is needed to stop covert smuggling of nuclear materials into the US. It is also important to keep "dirty" bombs away from attractive targets of opportunity. Detection of nuclear material is challenging. This is due both to radiation signals following a Poisson distribution and background radiation being ubiquitous. We discuss current approaches for reliable detection/ localization of radiation sources within acceptable false alarm rates; 2. Distributed vehicle behaviors - Self-driving cars are no longer science fiction. Applications, such as collision avoidance and platooning, posit interactions between multiple vehicles that are owned and maintained by more than one entity. To avoid disaster, what assumptions can be made when designing and implementing these behaviors? To make the system robust, it is best to make no assumptions. We explain design principles for implementing a platooning system that functions well, even when interacting with poorlymaintained vehicles. 3. The electric grid - creating an effective feedback loop can make the electric grid more efficient and able to include renewable power sources like wind and solar. Synchrophasor sensors send real-time information to power gird control centers. These network feeds are secured using virtual private networks to prevent attackers from manipulating sensor signals. We explain how these security mechanisms are vulnerable to disruption. We also consider how these vulnerabilities are inherent to the current IP network design. We give an overview of current challenges in the design and deployment of cyber-physical systems.