A trustworthiness evaluation framework for distributed networks

This research examines the utility of Markov switching models in assessing trust and trustworthiness of a heterogeneous network, e.g. distributed sensor networks. As an unsupervised machine learning method, hidden Markov models (HMM) is independent of the assumptions commonly used in modeling trust in complex systems. A relevant time series that switches regimes from trusted to untrusted periods of times is simulated to illustrate the theory of HMM and its effectiveness in Trust modeling. In this paper, we have employed HMM to estimate the parameters of a unified trust model that could make continual determinations of the trustworthiness of the data collected in any application environment. The results indicate that this method could effectively accommodate the desired features of our specified trust model despite various noises and uncertainties in the input signal. This study, by defining a new metric of trustworthiness and using HMM, provides an improvement over past studies in terms of computation costs, accuracy of estimation and forecasting, less a priori assumptions, and system agnosticism.