Remote attestation is an essential feature of Trusted Computing that allows a challenger to verify the trustworthiness of a target platform. Existing approaches towards remote attestation are largely static or too restrictive. In this paper, we present a new paradigm in remote attestation that leverages recent advancements in intrusion detection systems. This new approach allows the modeling of an application’s behavior through stochastic models of machine learning. We present the idea of using sequences of system calls as a metric for our stochastic models to predict the trustworthiness of a target application. This new remote attestation technique enables detection of unknown and zero-day malware as opposed to the known-good and known-bad classification currently being used. We provide the details of challenges faced in the implementation of this new paradigm and present empirical evidence supporting the effectiveness of our approach.