Robust and Energy Efficient Hardware-Oriented Security for IoT Systems and Applications

Internet of Things (IoT) is a revolutionary network that is envisioned to connect physical entities to the cyber world. IoT technology has fundamentally changed how we interact with our world. Worldwide spending on IoT is forecast to reach $745 Billion in 2019, and it is expected that investments in the technology will maintain double-digit growth rate for years to come.Despite wide adoption and strong anticipation in the technology, two major obstacles heavily constrained the further development in IoT, respectively security and energy challenges. From the security perspective, the entire lifecycle of an IoT device could potentially be vulnerable to various types of attacks. Since many devices are deployed in an insecure environment, attackers could gain unauthorized access to the exposed hardware, which invalidates many security assumptions made in traditional security research. From the energy perspective, many IoT devices are incapable of affording traditional cryptographical protection due to low energy and computation budget. Energy efficiency is therefore crucial for designs and establishments of IoT.To address IoT security problems, we explore and propose novel hardware-oriented security primitive designs and optimization techniques in this thesis. We first investigate the vulnerabilities of physically unclonable function (PUF), a popular low power hardware security primitive used in IoT devices, through the creation of a hardware emulation platform using programmable delay lines (PDL). To address vulnerabilities in PUFs, we propose a novel security primitive, Interconnected PUF Network (IPN), that interconnects small segments of strong PUFs in a reconfigurable network, limiting the single-bit prediction accuracy to as low as 53.19% against a wide range of modeling attacks. We demonstrated that the interconnections in an IPN can be optimized to maximize output randomness and stability using our proposed evolution-strategies-based algorithm. Looking beyond PUF-based security, we designed content-driven injective functions (CRIF) that rearrange compositions of hardware injective functions based on previous messages, providing secure message encryption/decryption between IoT devices. Facing the energy challenges, we propose ``computing while racing" technique that reduces 40.4% of area overhead and 7.69% of power when implementing arbiter PUF and arbitrary logic on field-programmable-gate-arrays (FPGAs). This is achieved through encoding digital signals in analog forms and achieves a high percentage of hardware sharing, suggesting resource sharing could potentially be a promising direction for power/energy reduction in IoT devices. Eventually, we propose two practical IoT applications. We first design a device anomaly detection utilizing the inconsistency in environmentally sensitive PUF challenge-response pairs. We show that our detector is more flexible and more power-efficient compared to state-of-the-art system monitors. Secondly, we demonstrate that our proposal of PUF-assisted group key management protocol securely protects IoT group communications while reducing global energy consumption by 47.3% compared to cryptographic key management solutions