Malicious users may reprogram cognitive radios to disregard how their actions may harm other users. To cope with such threats, we propose that cognitive radios must have a trusted domain, which contains an onboard regulation mechanism (ORM) that 'regulates' the radio's behavior by monitoring the radio, and taking appropriate action. We present a scheme for monitoring the link quality, in real-time to assess the proper action to take. Our method promptly prevents misbehavior by taking action in a manner proportional to the level by which the cognitive radio deviates from normal, allowed behavior. We demonstrate a method for the ORM to collect the link state information needed to regulate its cognitive radio. Lastly, we examine the resulting algorithms and study our regulation relative to traditional CSMA and AlohaMAC protocols on the ORBIT testbed using USRP/GNU Radios.