Machine Learning-driven Trust Prediction in Mobile Edge Computing-based Internet of Things Systems

Mobile Edge Computing (MEC) based Internet of Things (IoT) is a newly emerging computing infrastructure, which is being rapidly adopted in many real-world applications. Characterised by the pooled computing and storage resources co-located in geographically dispersed base stations of cellular networks, MEC allows IoT services to be deployed at the edge of the network providing faster access to their consumers. Such a system architecture also enables high-volume and high-velocity IoT data to be processed within MEC environments closer to where the data originates. This helps reduce the network stress on the core networks of mobile network providers considerably. In the meantime, deployed in massive numbers and different shapes as well as granularities, the aforementioned IoT services provide useful functionalities to interested service consumers in close proximity. Trust in such a system is an indispensable element, which enables secure interactions between IoT services and their consumers. It also assists people and smart devices alike to consume useful IoT services with an elevated degree of confidence that these interactions will bring favourable outcomes. For instance, in fitness tracking where the personally identifiable information such as social profile, behavioural and location data is shared with third party services, trust provides assurance that this collected information will be used as agreed by all parties. In intelligent transport systems, trust allows autonomous vehicles to determine services that provide credible location and traffic information supplied by vehicular networks, etc. As such, trust can be deemed to play an integral part towards ensuring user acceptance towards IoT systems. In this research, we propose a distributed machine learning architecture to effectively predict trustworthiness of sensor service providers in MEC-based IoT systems. The proposed architecture takes a predictive approach to evaluate the trustworthiness of services deployed in an inherently distributed MEC-based IoT environment taking into account the holistic characteristics of trust, also adhering to the systems characteristics of MEC. To that end, the aforementioned architecture models the task of training a trust prediction model over a distributed family of MEC environments within a given MEC topology as a distributed trust prediction problem. It then attempts to solve the aforesaid formulation collaboratively and in parallel using variants of Alternating Method of Multipliers (ADMM) under various circumstances led by how trust information is generated, collected and processed. The novelty of our work predominantly stems from our holistic approach to evaluate trustworthiness of IoT services in the following settings. We first proposed a data-driven distributed machine learning approach to scalably predict the trustworthiness of homogeneous IoT services in heterogeneous MEC-based IoT systems. The proposed approach formulates training distributed trust prediction models within an MEC-based IoT system as a Network Lasso problem. To efficiently and effectively tackle the high-volume silos of trust information generated within each MEC environment, we then introduced a variant of the Stochastic Alternating Method of Multipliers (S-ADMM) framework enriched with the ability for feature selection at each MEC layer. Furthermore, we observed that MEC-based IoT systems generate trust information in a distributed manner in real-time. More importantly, it is imperative to investigate how this real-time trust information could be effectively integrated into trust prediction strategies in order to capture the ever evolving nature of trustworthiness of IoT services. In turn, such a strategy allows IoT service consumers to derive more relevant and accurate trust decisions. To that end, we also proposed an online approach to train a family of distributed trust prediction models in a given MEC topology. We then adopted the Online Alternating Direction Method (O-ADM) to effectively train trust prediction models in parallel over each distributed MEC environment. Moreover, we also proposed a data-driven and context-aware approach to bootstrap trustworthiness of homogeneous IoT services in MEC-based IoT systems. The proposed approach addresses key limitations in adapting existing trust bootstrapping approaches into MEC-based IoT systems. These key limitations include, the lack of opportunity for a service consumer to interact with a lesser-known service over a prolonged period of time to get a robust measure of its trustworthiness, inability of service consumers to consistently interact with their peers to receive reliable recommendations of the trustworthiness of a lesser-known service as well as the impact of uneven context parameters in different MEC environments causing uneven trust environments for trust evaluation. In addition, the proposed approach also tackles the problem of data sparsity via enabling knowledge sharing among different MEC environments within a given MEC topology.