Power Analysis on Error Correcting Codes


#1

Hi, I’m a litlle bit familiar with power analysis on AES but I’m completely new to the concept of power analysis on Error Correcting Codes (ECC). I tried to do CPA on Extended Golay (24,12,8) ECC and this is how the procedure looked like:

  1. Loaded the immplementation (code written in VHDL) onto SASEBO-GIII and there is a randomly generated noisy 24-bit word that comes as an input.
  2. After the trigger, the noisy word gets corrected into a valid codeword. Like this, I recorded 15000 traces for randomly generated (used randi() function in matlab) 15000 noisy words. [Sample rate:2.5GS/s]
  3. Since this a white-box experiment, I can see both inputs and valid codewords (outputs). When I tried to plot the correlation between all 15K hamming distances between inputs and outputs and the 15K traces corresponding to it, it ran along the time domain.

However, I was not able to figure out a method in which I can guess the valid codeword after getting corrected with the noisy words and the traces in hand.

If anyone could help me with some suggestions, I’d be very much grateful.

Also, please do let me know if this lacks clarity so that I can improvise it accordingly!

Thanks in advance! :slight_smile: