Device authentication and identification has been recently cited as one of the most pressing security challenges facing the Internet of things (IoT). In particular, the open-access nature of the IoT renders it highly susceptible to insider attacks. In such attacks, adversaries can capture or forge the identity of the small, resource constrained IoT devices and, thus, bypass conventional authentication methods. Such attacks are challenging to defend against due to the apparent legitimacy of the adversaries' devices. The primary goal of this research is to overcome this challenge by developing new authentication methods that supplement traditional security solutions with cyber-physical fingerprints extracted from the IoT devices' environment. This project will develop a novel machine learning framework that enables the IoT to dynamically identify, classify, and authenticate devices based on their cyber-physical environment and with limited available prior data. This will result in the creation of environment-based IoT device credentials that can serve as a means of attestation, not only on the legitimacy of a device's identity, but also on the validity of the physical environment it claims to monitor and the actions it claims to be performing over time. The framework will also encompass an experimental IoT software platform that will be built to validate the proposed research. Owing to a partnership with the NIST Global City Teams Challenge (GCTC) project "Bringing Internet of Things Know-How to High School Students", a collaboration with IoT-DC, Arlington County, VA, and other entities, the proposed research will train high school students, STEM educators, and a broad community on a variety of research topics that will include IoT security, cyber-physical systems, and data analytics. The broader impacts will also include the creation of an interdisciplinary workforce focused on securing tomorrow's smart cities.