With the massive deployment of mobile devices in sensitive areas such as healthcare and defense, hardware reliability and security have become hot research topics in recent years. These topics, although different in definition, are usually correlated. This dissertation offers an in-depth treatment on enhancing the reliability and security of low power circuits and systems. The first part of the dissertation deals with the reliability of sub-threshold designs, which use supply voltage lower than the threshold voltage (V th) of transistors to reduce power. The exponential relationship between delay and Vth significantly jeopardizes their reliability due to process variation induced timing violations. In order to address this problem, this dissertation proposes a novel selective body biasing scheme. In the first work, the selective body biasing problem is formulated as a linearly constrained statistical optimization model, and the adaptive filtering concept is borrowed from the signal processing community to develop an efficient solution. However, since the adaptive filtering algorithm lacks theoretical justification and guaranteed convergence rate, in the second work, a new approach based on semi-infinite programming with incremental hypercubic sampling is proposed, which demonstrates better solution quality with shorter runtime. The second work deals with the security of low power crypto-processors, equipped with Random Dynamic Voltage Scaling (RDVS), in the presence of Correlation Power Analysis (CPA) attacks. This dissertation firstly demonstrates that the resistance of RDVS to CPA can be undermined by lowering power supply voltage. Then, an alarm circuit is proposed to resist this attack. However, the alarm circuit will lead to potential denial-of-service due to noise-triggered false alarms. A non-zero sum game model is then formulated and the Nash Equilibria is analyzed.