Attacking Acros MixColumns

Embedded Security
1

Hello, I encountered some problems while using the Attacking Acros MixColumns algorithm to attack AES keys. I hope to receive your help. Thank you very much.
The problem description is as follows:
Background: When using the CW1173 device to attack a custom target version of AES key, the sampling rate was set to 100MHz, resulting in a low success rate of the attack. Generally, about 3 out of 16 bytes in the result are incorrect. We collected 1000 trajectories for each calculation, and each trajectory had 20 peak signals during AES operation.
Problem point:

  1. How to align multiple trajectories collected. For example, is it alignment of the starting point of the trajectory, alignment of the peak during each AES operation, or any other alignment method???
  2. What filtering algorithm would be better for filtering the collected trajectory data?? How to determine if the filtered data is appropriate???
  3. In a single trajectory data, there may be slight differences in the time interval between two AES operation peaks, around 50ns. Does this deviation affect the success rate??
2

Is “trajectory” a power trace? Have you tried simply increasing the number of power traces?

It’s hard for me to answer your questions with certainty because I don’t know anything about your custom target and firmware.

Traces should be aligned where the leakage is present. The hard part is finding that. If you have control over the target firmware then it’s easier because you can simply move the trigger_high() to the start of the MixColumn step.

There is no universal answer to that, it very much depends on your target.

Maybe! As I said what matters most is that the leakage be at a constant place in time.