I was trying to understand the procedure of the DPA attack on Stream cipher (like Mickey-128).
I learned that, the attacker need to build hypothetical power consumption matrix with different initial vectors. For example: using hamming-distance model to describe the power consumption based on R & S registers. But by doing this, different initial vectors have to be known to the attacker.
So my question is, in real life, how the attacker get these large amount of different initial vectors to build the hypothetical power consumption matrix?