P14_21: Design Obfuscation and Performance Locking Solutions for Analog/RF ICs
Topic Areas: Security and Trust Solutions for Analog/Mixed-Signal/RF Circuits
Principal investigator: Dr. Yiorgos Makris, University of Texas at Dallas
Co-Principal investigator(s): Dr. Ken O, Dr. Andrew Marshall, University of Texas at Dallas
PI Email
Abstract:
This multi-year effort seeks to develop design obfuscation and performance locking solutions for analog/RF Integrated Circuits (ICs). Unlike their digital counterparts, for which a plethora of logic locking and design obfuscation solutions have been developed in the past, the analog/RF domain has not yet received the same level of attention, mainly because the problem is fundamentally different in this domain. Indeed, digital design obfuscation relies on the sheer number of logic gates and the discrete nature of Boolean functions. Analog/RF designs, however, consist of a fairly small number of transistors interconnected in a fairly small number of known topologies for each type of component. Therefore, obfuscating the functionality of an analog/RF design is, most likely, an ineffective proposition. Instead, efforts should focus on hiding the performance of an analog/RF IC, by obfuscating the parameters of a design, as this is the true intellectual property (IP) which the designers spend the vast majority of their time on. To this end, this multi-year project seeks to investigate a variety of solutions for protecting the IP of an analog/RF design, by obfuscating the actual design parameters and by locking the ability to calibrate a fabricated IC within its intended specifications. During the first year of the project, we have already implemented and demonstrated an analog performance calibration locking method through the use of cyclically-connected analog floating gate transistors, which we demonstrated on an Operational Transconductance Amplifier (OTA). Additionally, we are in the process of developing a similar method based on a digitally-programmable selection of minimum-sized transistor pairs, which we are using to minimize phase noise of a 6GHz Voltage Controlled Oscillator (VCO). In the second year of the project, we propose to pursue a novel idea that we have recently come up with for addressing this problem in a design-independent and generalizable way, namely distributed parametric obfuscation. Additionally, through resources that we have secured outside the auspices of the NSF CHEST I/UCRC, we propose to fabricate a custom integrated circuit which will showcase all of the above solutions in GlobalFoundries’ 130nm technology and will corroborate their cost-effectiveness in analog in silicon.