Text dependent speaker verification using MFCC and vector quantization

Question

I am trying to implement a text dependent speaker verification system. Here each speaker will speak their unique identification number and they will be granted access. I am trying to implement this using MFCC features and Vector Quantization. Till now the system is performing good. However, the system has become text-independent. No matter what a speaker speaks( even if he speaks the ID number of another), he gets identified. But I want my system to detect when a user is trying to impersonate another user. How can I do this using signal processing?