Approaching bandwidth reliability of a CMOS highly dense physical unclonable function

A Physical Unclonable Function (PUF) is a random cryptographic function which could be implemented using a complex process-based physical system. PUF is a good inexpensive candidate for producing signature keys for cloud-based reliable communication within Internet of Things (IoT) devices in future. PUF made from on-chip Ring Oscillator (RO) is popular due to random physically induced frequency behavior and known as RO-PUF. Frequency bandwidth of applied ROs are subject to change due to environmental or aging variation, this results in bit flip and consequently non-reliable behavior of the RO-PUF. A frequency spectrum analysis and correction for the RO-PUF based on an industrially implementable and inexpensive hardware technology is required. Most of previous works are either software level works which do not offer a silicon-level reliability view, or are based on Field Programmable Gate Array (FPGA) which are not suitable for fabrication of the low to middle price range PUF. In this paper, we have presented a statistical method for improving reliability of PUF designs. We have introduced RO-PUF Bandwidth Reliability Measure “BRM”. BRM is improved by a 31.2 times scaling factor via applying a hardware level tuning on a highly dense RO-PUF.