Spectrum Enforcement and Liability Assignment in Cognitive Radio Systems

The advent of frequency-agile radios holds the potential for improving the utilization of spectrum by allowing wireless systems to dynamically adapt their spectral footprint based on the local conditions. Whether this is done using market mechanisms or opportunistic approaches, the gains result from shifting some responsibility for avoiding harmful interference from the static "regulatory layer" to layers that can adapt at runtime. However, this leaves open the major problem of how to enforce/incentivize compliance and what the structure of "light-handed" regulation should be. This paper examines this and focuses on two specific technical problems: (a) determining whether harmful interference is occurring and (b) assigning liability by detecting the culprits. "Light-handed regulation" is interpreted as making unambiguous (and easily certified) requirements on the behavior of individual devices themselves while still preserving significant freedom to innovate at both the device and the system level. The basic idea explored here is to require the PHY/MAC layers of a cognitive radio to guarantee silence during certain time- slots where the exact sequence of required silences is given by a device/system -specific code. Thus, if a system is a source of harmful interference, the interference pattern itself contains the signature of the culprit. Nevertheless, identifying the unique interference pattern becomes challenging as both the number of cognitive radios and the number of harmful interferers increases. The key tradeoffs are explored in terms of the "regulatory overhead" (amount of enforced silence) needed to make guarantees. The quality of regulatory guarantees is expressed by the time required to convict the guilty, the number of potential cognitive systems that can be supported, and the number of simultaneously guilty parties that can be resolved. We show that the time to conviction need only scale logarithmically in the potential number of cognitive users. The bas- - e of the logarithm is determined by the amount of overhead that we will tolerate and how many guilty parties we want to be able to resolve.