In recent decades, the smart cities are incorporating with Internet-of-Things (IoT) infrastructures for improving the citizens’ quality of... Show moreIn recent decades, the smart cities are incorporating with Internet-of-Things (IoT) infrastructures for improving the citizens’ quality of life by leveraging information/data. The huge amount of data is extracted and generated from the devices (e.g., mobile applications, GPS navigation systems, urban traffic cameras, etc.), or city sectors such as Intelligent Transportation Systems (ITS), Resource Allocation, Utilities, Crime Detection, Hospitals, and other community services.This dissertation aims to systematically research the Data Analysis in IoT System, which mainly consists of two aspects: Utility and Efficiency. First, ITS as a representative system in IoT in the smart city, I present the work on privacy preserving for the trajectories data, which is achieved by the differential privacy technique with a novel sanitation framework. Moreover, I have studied the resource allocation problem in two different approaches: Cryptographic computation and Hardware en- claves with the utility and efficiency accordingly. For the Cryptographic computation approach, I utilize Secure Multi-party Computation (SMC) technique for achieving the privacy-aware divisible double auction without a mediator. Besides, I also pro- pose a hardware-based solution Trusted Execution Environment (TEE) for performance improvement. At the same time, integrity and confidentiality are also able to be guaranteed. The proposed hybridized Trusted Execution Environment (TEE)- Blockchain System is designed for securely executing smart contract. Finally, I have studied the Cryptographic Video DNN Inference for the smart city surveillance, which privately inferring videos (e.g., action recognition, and video, and classification) on 3D spatial-temporal features with the C3D and I3D pre-trained DNN models with high performance. This dissertation proposes the privacy preserving frameworks and mechanisms are able to be applied efficiently for IoT in the real-world. Show less
