This paper addresses the fingerprinting of communication protocols based on temporal and behavioral information. The objective of fingerprinting a device speaking a given protocol is to uniquely identify the device by looking at captured traffic that is generated by devices implementing that protocol. This paper proposes a conceptual model for capturing behavior and related temporal information from devices that implement a given protocol. Our key contribution is a fingerprinting scheme, where individual fingerprints are represented by tree-based temporal finite state machines. We have developed a fingerprinting scheme that leverages supervised learning approaches based on support vector machines for this purpose. We have validated the proposed approach on Session Initiation Protocol and concluded that very good classification performance is achieved.