This paper presents an enhancement of the accuracy and security of biometric templates based on a code-based cryptosystem—McEliece cryptosystem—which in addition to randomness is also probabilistic, which provides higher susceptibility of templates towards brute force attacks. It is possible to generate many different secure biometric templates for the same system and also unique biometric templates for multiple systems from the same biometric trait; it is just a matter of using a different error vector. It is also easy to cancel a secure template by simply deleting the compromised template and generating a new one by using different error vector.