Detecting Impersonation Attacks in a Static WSN

Abstract: The current state of security found in the IoT domain is highly flawed, a major problem being that the cryptographic keys used for authentication can be easily extracted and thus enable a myriad of impersonation attacks. In this MSc thesis a study is done of an authentication mechanism called device fingerprinting. It is a mechanism which can derive the identity of a device without relying on device identity credentials and thus detect credential-based impersonation attacks. A proof of concept has been produced to showcase how a fingerprinting system can be designed to function in a resource constrained IoT environment. A novel approach has been taken where several fingerprinting techniques have been combined through machine learning to improve the system’s ability to deduce the identity of a device. The proof of concept yields high performant results, indicating that fingerprinting techniques are a viable approach to achieve security in an IoT system.