With the confluence of cutting-edge machine learning algorithms and highly efficient computing

hardware, IoT-enabled edge intelligence has revolutionized a myriad of applications from health care, wearable devices, industrial manufacturing, to social networking, virtual assistant, and security. The primary operation of these applications is to exchange and process a large volume of data, driving the need for higher memory need and bandwidth. Memory circuits are performance-critical in various IC chips such as microprocessors, FPGA, DSP, ASICs, and SoCs. On the other hand, memory circuits and data storage systems are vulnerable to hardware attacks and can easily become the target of untrusted hardware due to their mass production by numerous vendors. Memory circuits are typically built upon small transistors to achieve high-density integration, but small transistors are sensitive to many attacks such as supply voltage glitching. Technology scaling reduces the capacitance, supply voltages, and hence the information-bearing charges at storage nodes. Consequently, fault injection attacks can easily flip memory bits. Furthermore, memory read and write accesses often rely on preciously tuned clock signals to perform latching, gating, dynamic timing, and sophisticated cycle borrowing functions. As the size and access speed of memory continue to increase, memory circuits

become increasingly vulnerable to clock glitching attacks. Thus, there is an urgent need to develop effective security monitoring and attack detection techniques to ensure data integrity on even untrusted hardware platforms.