A robust and secure database for spectrum sharing in cognitive radio networks that is obscured from the viewpoint of secondary users is presented. The database allocations secure features of white space resource usage from being learned. The design of non-inferable database is based on two cases. In the first case, the primary or spectrum lender has no knowledge of secondary users or potential jammers among them. In this case, the problem is modeled as a Markov Decision Process. In the second case, the primary system has some knowledge about spectrum borrowers and the problem is modeled as a Bayesian Stackelberg game.
Mutual information interpretation of the Bayesian Stackelberg game is also presented. The solutions facilitate releasing more bandwidth as required by US National Broadband Plan. Applications of this scheme are manifold. This design can be used for securing spectrum resources, for example radar white spaces, while being shared with LTE and commercial communication systems. Further, it provides jammer-proof spectrum sharing among various communication, detection, and navigation systems. Simulation results verify this scheme improves system throughput while maintaining desired obfuscation level or entropy.