CMOS-compatible RRAM-based structures for the implementation of Physical Unclonable Functions (PUF) and True Random Number Generators (TRNG)

点击翻译按钮获取中文摘要

Physical Unclonable Functions (PUF) and True Random Number Generators (TRNG) are two components widely used nowadays to generate random bit streams in security applications. The huge increase in the last decade in the use of portable consumer electronics has revealed the security in wireless communications as one of the most important requirements to be fulfilled in microelectronics technology. Therefore, it is of great interest to develop an implementation of these components which accomplishes the following characteristics: low-power operation, high-integration density, and compatibility with CMOS processes. Following the “More than Moore” approach (Increase of the performance adding functionality) such characteristics can be achieved. For instance, Resistive Random Access Memories (RRAM) have emerged in the last years as promising candidates in the field of Non-Volatile Memories (NVM). Moreover, the mechanisms behind switching operations in RRAM devices are intrinsically stochastic. Therefore, RRAM technology has started recently to be considered as a suitable solution to implement the future PUF and TRNG components. The study proposed in this project involves interdisciplinary research in order to achieve three main targets:1. Studying in detail the statistical distributions of the electrical parameters involved in RRAM switching, which have been typically used as a source of randomness.2. Figuring out how the correlations which avoid the true randomness emerge from fundamental physical and chemical processes.3. Development of an appropriate operative algorithm able to overcome the correlations found on the electrical parameters of RRAM devices providing the true random digital outputs required for both TRNG and PUF applications. In order to understand why the electrical characteristics of RRAM devices are intrinsically stochastic but not true random, a complete materials study and electrical characterization will be performed with these devices. The RRAM-based structures required for this characterization will be fabricated by starting from the well-known TiN/HfO2/Ti/TiN structure. The fabrication parameters will be modified to assess their influence in the randomness. To link electrical characteristics to physical and chemical atomic interactions, a complementary approach is required: the simulation of atomistic models. Finally, the statistical analysis will be crucial to guide the design of the operative algorithm for the implementation of PUF aa well as TRNG.