Having confidence that a user can identify a machine they want to communicate with is critical for users who wish to keep their data safe. Being able to establish the identity of a machine is of critical importance for establishing user trust. In this paper, we outline a methodology to leverage the ubiquitous USB interface and power measurements consisting of voltage, wattage, and amperage to uniquely identify machines. We collect USB power samples from a corpus of 45 machines on the University of Oregon campus and through machine learning classifier techniques, we demonstrate that it is possible to use statistical metrics extracted from these samples to differentiate hosts based on class label. Using these statistical metrics, we are able to correctly differentiate sampled hosts by ID with an upwards of 94% accuracy. In addition we are able to identify characteristics about the machines such as model number and operating system with an accuracy of 98%. Using a Random Forest classifier were are able to generate fingerprints that are capable of consistently distinguishing hosts that are seemingly identical with an accuracy of 98% as well. Later in our analysis we show that our methods can be extended to determine other characteristics about target hosts as well such as what USB devices are connected. Our techniques are deployable in an accessible, low-cost fashion.