Electronic Thesis and Dissertation Repository

Thesis Format

Monograph

Degree

Doctor of Philosophy

Program

Electrical and Computer Engineering

Supervisor

Wang, Xianbin

Abstract

The fifth-generation (5G)-and-beyond networks will provide broadband access to a massive number of heterogeneous devices with complex interconnections to support a wide variety of vertical Internet-of-Things (IoT) applications. Any potential security risk in such complex systems could lead to catastrophic consequences and even system failure of critical infrastructures, particularly for applications relying on tight collaborations among distributed devices and facilities. While security is the cornerstone for such applications, trust among entities and information privacy are becoming increasingly important. To effectively support future IoT systems in vertical industry applications, security, trust and privacy should be dealt with integratively due to their close interactions. However, conventional technologies always treat these aspects separately, leading to tremendous security loopholes and low efficiency. Existing solutions often feature various distinctive weaknesses, including drastically increased latencies, communication and computation overheads, as well as privacy leakage, which are extremely undesirable for delay-sensitive, resource-constrained, and privacy-aware communications.

To overcome these issues, this thesis aims at creating new multi-dimensional intelligent security provisioning and trust management approaches by leveraging the most recent advancements in artificial intelligence (AI). The performance of the existing physical-layer authentication could be severely affected by the imperfect estimate and the variation of physical link attributes, especially when only a single attribute is employed. To overcome this challenge, two multi-dimensional adaptive schemes are proposed as intelligent processes to learn and track the all available physical attributes, hence to improve the reliability and robustness of authentication by fusing multiple attributes. To mitigate the effects of false authentication, an adaptive trust management-based soft authentication and progressive authorization scheme is proposed by establishing trust between transceivers. The devices are authorized by their trust values, which are dynamically evaluated in real-time based on the varying attributes, resulting in soft security and progressive authorization. By jointly considering security and privacy-preservation, a distributed accountable recommendation-based access scheme is proposed for blockchain-enabled IoT systems. Authorized devices are introduced as referrers for collaborative authentication, and the anonymous credential algorithm helps to protect privacy. Wrong recommendations will decrease the referrers’ reputations, named as accountability. Finally, to secure resource-constrained communications, a lightweight continuous authentication scheme is developed to identify devices via their pre-arranged pseudo-random access sequences. A device will be authenticated as legitimate if its access sequences are identical to the pre-agreed unique order between the transceiver pair, without incurring long latency and high overhead.

Applications enabled by 5G-and-beyond networks are expected to play critical roles in the coming connected society. By exploring new AI techniques, this thesis jointly considers the requirements and challenges of security, trust, and privacy provisioning, and develops multi-dimensional intelligent continuous processes for ever-growing demands of the quality of service in diverse applications. These novel approaches provide highly efficient, reliable, model-independent, situation-aware, and continuous protection for legitimate communications, especially in the complex time-varying environment under unpredictable network dynamics. Furthermore, the proposed soft security enables flexible designs for heterogeneous IoT devices, and the collaborative schemes provide efficient solutions for massively distributed entities, which are of paramount importance to diverse industrial applications due to their ongoing convergence with 5G-and-beyond networks.

Summary for Lay Audience

The fifth-generation (5G)-and-beyond networks will provide broadband access to a massive number of heterogeneous devices with complex interconnections to support a wide variety of vertical Internet-of-Things (IoT) applications. Any potential security risk in such complex systems could lead to catastrophic consequences and even system failure of critical infrastructures. While security is the cornerstone for such applications, trust among entities and information privacy are becoming increasingly important. To effectively support future IoT systems in vertical industry applications, security, trust and privacy should be dealt with integratively due to their close interactions. However, conventional technologies always treat these aspects separately, leading to tremendous security loopholes and low efficiency. Existing solutions often feature various distinctive weaknesses, including drastically increased latencies, communication and computation overheads, as well as privacy leakage, which are extremely undesirable for delay-sensitive, resource-constrained, and privacy-aware communications.

To overcome these issues, this thesis aims at creating new multi-dimensional intelligent security provisioning and trust management approaches by leveraging the most recent advancements in artificial intelligence (AI). Two multi-dimensional adaptive schemes are proposed as intelligent processes to learn and track the all available physical attributes, hence to improve the reliability and robustness of authentication by fusing multiple attributes. To mitigate the effects of false authentication, an adaptive trust management-based soft authentication and progressive authorization scheme is proposed by establishing trust between transceivers. By jointly considering security and privacy-preservation, a distributed accountable recommendation-based access scheme is proposed for blockchain-enabled IoT systems. Finally, to secure resource-constrained communications, a lightweight continuous authentication scheme is developed to identify devices via their pre-arranged pseudo-random access sequences.

By exploring new AI techniques, this thesis jointly considers the requirements and challenges of security, trust, and privacy provisioning, and develops multi-dimensional intelligent continuous processes for ever-growing demands of the quality of service in diverse applications. These novel approaches provide highly efficient, reliable, model-independent, situation-aware, and continuous protection for legitimate communications. Furthermore, the proposed soft security enables flexible designs for heterogeneous IoT devices, and the collaborative schemes provide efficient solutions for massively distributed entities, which are of paramount importance to diverse industrial.

Download

Share

COinS