A Hidden Markov Model security scheme for query state inference in discovery services

Discovery Services refer to a suite of network services enabling efficient track-and-trace capabilities of objects in the Internet of Things (IoT). Deployment of such services may be performed in the form of simple queries originating from the corresponding stakeholders either to store/retrieve data in/from the Cloud. An example of such services is the EPCglobal Discovery Services. The extremely sensitive nature and the expected large scale of the exchanged data in the IoT (e.g, the EPCglobal Network) highlight the importance of a security scheme capable of distinguishing safe queries from risky ones, based both on a vector of observed real values extracted from the current query, and on a pattern inferred from the past queries. In this paper, we propose a probabilistic security scheme enhancing the accuracy of detecting risky queries in the EPCglobal Network. Our proposed scheme is based on a Hidden Markov Model (HMM) which is first trained, then used to infer the state of the query at hand. We assume that the observed real values, extracted from the queries, follow Gaussian distributions, depending on the inherent nature of the query at hand; i.e. safe or risky. We conducted extensive experiments. The results show that our HMM-based security scheme enhances the accuracy of detecting risky queries.