Privacy Modelling and Preservation for Assisted Living within Smart Homes




Journal Title

Journal ISSN



Volume Title


De Montfort University


Thesis or dissertation

Peer reviewed


In the last decades, the increase of the ageing population worldwide has created a need for Ambient Assisted Living (AAL) solutions that allow people to continue living and ageing in their own homes comfortably and safely. AAL solutions typically combine IoT technologies and machine learning to provide services that are context-aware and personalised. However, modern machine learning and internet of things systems present new challenges in privacy and security. With the collection of large datasets and increasingly complex analysis tasks, privacy has become a critical issue in these systems. This thesis aims to understand users’ privacy concerns and preferences and develop mechanisms to allow them better control over their privacy in the context of AAL and Smart Homes. In more detail, it presents the design, development and evaluation of privacy preserving methods. The research has been conducted through four main studies. The first one aims to understand users’ notion of privacy in connection with IoT and the contextual factors that affect it. The second study investigates users’ privacy risk, ways to manage it, and how such metrics could be embedded into the design of modern intelligent negotiation systems to raise awareness and better protect privacy. The third study examines anonymisation and data sharing within Ambient Assisted Living and develops methods to provide privacy-preserving machine learning with differential privacy. Lastly, the fourth study, proposes a privacy-preserving architecture for AAL, combining all the aforementioned studies into an end to end system. This thesis implemented privacy-preserving mechanisms for different machine learning tasks, as well as utility tools and AAL data simulators, for experimentation. All the proposed methods have been evaluated using various real and synthetic datasets. The experimental results of this dissertation demonstrate that it is feasible to design efficient privacy-preserving machine learning systems with negligible costs in utility and performance. Additionally, the lessons learned in this dissertation may inform and assist service providers and developers in designing practical end to end privacy-preserving architectures in emerging areas such as privacy-preserving machine learning and internet of things.





Research Institute