Skip to main content
Open Access Publications from the University of California

Building trust into light-handed regulations for cognitive radio

  • Author(s): Woyach, Kristen Ann
  • Advisor(s): Sahai, Anant
  • et al.

This thesis introduces an incentive-based trust model to let wireless spectrum regulation embrace diverse current and future means of implementing cognitive radio.

Cognitive radio has emerged as a way to combat inefficient spectrum use by allowing independently designed networks to share the same frequency band. This philosophy has been embraced by the FCC, which has already allowed cognitive use in the TV bands, and plans to make spectrum sharing the norm in other bands as well. To enact spectrum sharing, regulatory decisions, like band assignment, are made at runtime so that they can reflect local context. From a regulatory perspective, the most important question is how to trust that these decisions will be made and carried out correctly.

Right now, the FCC guarantees correct decisions by directly testing that any deployed technologies are incapable of making bad decisions. This process of testing is called certification. But certification has limitations. For example, a network of nodes could sense for a TV signal and decide as a group that the TV tower is far enough away that their interference to TV receivers would be negligible. However, this network will never pass a certification test. There is no way to prove that the network will stay silent if all the nodes are blocked by the same building so they cannot sense the tower but can cause interference.

This thesis provides a new model for trust that would allow networked sensing and any other novel spectrum sharing solution through light-handed regulations. The idea is to build a system that allows regulators to trust secondaries to follow sharing rules regardless of whether they are technically capable of finding spectrum holes. This is accomplished by an incentive mechanism, a spectrum jail, that will punish secondaries caught causing interference by degrading their quality of service. This thesis shows that for such a mechanism to work, cognitive radio must be thought of as a band-expander. If the same mechanism must apply to all radios, regardless of technology, there must be pretty good unlicensed or licensed bands that secondaries can use if they cannot share spectrum appropriately.

The mechanism explored here is inspired by the ideas in the law and economics literature as well as the spectrum policy literature. This thesis takes these mostly rhetorical arguments and develops the first mathematical model for incentive-based trust in spectrum regulation.

This model allows identification of the most difficult to enforce cases: the regulator must decide whether a primary will be protected even if it hardly ever uses its band. The regulator must also decide what constitutes harmful interference. Some interference is unavoidable when bands are shared; the regulator must decide how much interference the primary must accept in a shared environment. When these decisions are made, this thesis shows that trust can be guaranteed with a sanction set at certification time and which is applied to all cognitive devices regardless of technology.

The model also gives quantitative performance metrics, measuring the ability for secondaries to reclaim spectrum holes, which illustrate the dependence on the regulator's ability to catch wrongdoers. In particular, this thesis shows that while trust depends on the ability to catch those causing interference, runtime performance depends on the wrongful conviction rate. So, even applying the same sanction, as spectrum sharing technology and catching technology improves, performance will improve as well.

This model is extended to understand what role the primary can or must play in its own protection as new primary devices are developed to operate in a shared band. By controlling the cost of reporting, the regulator can trust a primary to report interference correctly. This also means that if a secondary is difficult to coexist with, the primary will not use the jail system to try to get rid of the secondary. It will instead hire a ``band-sitter,'' which is a preferred secondary system that coexists more easily with the primary.

This thesis also addresses multiple secondaries and aggregate interference by giving a basic framework of results to guide research in this direction. The distribution of aggregate interference from randomly placed nodes is explored to understand placement risk: the threat of too much interference caused by clusters of secondaries too close to the primary. Then, the thesis develops strategies to use the secondary location information that TV whitespace databases already have to address the problem of placement risk. Finally, a basic queuing model is suggested as a future direction to extend spectrum jails to deal with multiple secondaries.

Finally, this thesis answers the question of why jails? The original motivation is two-fold. First, jails lend themselves to simple modeling because the utility and the sanction are both measured in quality of service terms. Second, jails can actually be reasonably implemented. The FCC has allowed TV whitespace devices using databases to coordinate spectrum access. In order to actually secure this operation, databases will need to be able to identify malfunctioning devices and turn them off. These same identity and kill-switch technologies will also enable spectrum jails. Jails can even be implemented through the databases themselves as a denial of operating tokens.

At a more philosophical level, in-kind and monetary sanctions are fundamentally different things. Which one is actually better suited to the spectrum sharing enforcement problem? The last chapter will apply the same performance-based understanding from the the rest of the thesis to understand when fines or in-kind punishments should be preferred. It shows that in cases of high uncertainty, or when primary protection is the most important consideration, in-kind sanctions are the right approach.

Main Content
Current View