Speaker:Hongyi Bian

Towards Trustworthy and Personalized Situation-aware Services in IoT-enabled and Learning-based Smart Systems

As we steadily advance into the era of situation-aware (or situ-aware) ubiquitous computing, the fast-growing technologies for the Internet of Things (IoT) have accelerated learning-driven approaches for synthesizing situations based on perceived contexts in smart environments. These IoT-enabled smart systems aim to improve the quality of life, mitigate potential hazards, and offer personalized services. However, challenges remain in the development of situ-aware ubiquitous computing towards personalized services, especially in the consideration of human factors necessary for realizing more fine-grained and trustworthy situ-aware services. On the one hand, the challenge of training a robust learning model is aggravated by the scarcity of locally collected data. Federated Learning (FL)-based approaches tend to address the issue in a centralized, top-down setting. Still, they fall short of facilitating personalized learning with heterogeneous local situation distributions. On the other hand, Blockchain-based IoT (BIoT) systems aim to mitigate trust-related risks often found in traditional cloud-based, vendor-specific IoT networks. However, the challenge of integrating learning-driven situation awareness with BIoT remains, primarily because of the conflicts between the deterministic nature of smart ledger operations and the non-deterministic nature of machine learning, as well as the high costs of conducting machine learning operations on blockchains. To address the challenges, we first proposed a framework named Situ-Oracle. With the framework, a computation oracle is leveraged to provide situation analysis (or situ-analysis) as a service to the local BIoT. Furthermore, we devised a learning-driven situ-analysis service with Recurrent Neural Network (RNN)-based models, and novel smart contracts are designed as intermediary communication channels between the IoT network to the computation oracle. Secondly, we studied a decentralized, clustering-based mutual-learning scheme. It allows each local smart system to obtain a personalized situ-analysis model via knowledge-sharing within the designated clusters in terms of similar situational distributions. We used connected smart homes as a case study to validate the feasibility and effectiveness of the framework design. Overall, the study presents a promising approach to enhanced trustworthiness and personalization of learning-driven situ-aware services in modern IoT-enabled smart environments.

Committee: Carl Chang (co-major professor), Wensheng Zhang (co-major professor), Simanta Mitra, Ying Cai, and Daejin Kim.

Join on Zoom: https://iastate.zoom.us/j/98079065596