Securing Cognitive Radio Networks with Dynamic Trust against Spectrum Sensing Data Falsification

The paper addresses the problem of spectrum sensing data falsification (SSDF) attacks in cooperative sensing and develops a dynamic trust management scheme to reliably detect and mitigate such attacks in cognitive radio networks. Secondary users (SUs) sense the spectrum to identify and dynamically access the available primary user (PU) channel. To improve sensing reliability, SUs send their sensing reports (subject to sensing error of individual spectrum sensor, e.g., Energy detector) to a fusion center, where sensing reports are combined to identify the channel state (idle or busy). Three types of SUs are considered: (i) normal users reporting their sensing results honestly, (ii) attackers reporting their sensing results inaccurately in form of random on off attacks, (iii) irrelevant users observing independent spectrum events. The consistency of sensing results is checked at the fusion center with channel outcome of scheduled transmissions (obtained under possible feedback errors) and the probabilistic trust of each user is updated with Bayes rule. Then, the optimal decision rule for cooperative sensing is determined dynamically with new trust assignments. Performance is evaluated separately under Gaussian signal assumption and with wireless transmission data collected with actual radios. Results show that dynamic trust management reliably detects user types and eliminates effects of SSDF attacks. Both the sensing error probability and the throughput are significantly improved compared to the case when all users are by default trusted to be normal users. This performance gain is maintained as the number of attackers or irrelevant users increases in the cognitive radio network.