A wireless sensor network consists of a number of sensors laid out in a field with mobile sinks dynamically aggregating data from the nodes. Sensitive applications such as military environment require the sink to identify if a sensor that it visits is legitimate, and in turn, the sensor has to ensure that the sink is authenticated to access its sensitive data. For the system to intelligently learn the credentials of non-malicious sink and non-malicious sensors based on the dynamically observed data, four approaches using access control lists, authenticator tokens, message digests, and elliptic curve variant of RSA algorithm are proposed along with the formal logic for correctness. The experimented data is analysed using false acceptance rate, false rejection rate, precision, and curve analysis parameters. The approaches are further compared based on the attacks they are vulnerable to and execution time, ultimately concluding that exchange of message digests and elliptic curve RSA algorithm are more widely applicable.