AES 256 Oscilloscope Probing on CW305 FPGA target

1

I am running AES256 in CW305. Probing the amplified output X4 from the board, I got the following traces:

image
image1148×781 86.9 KB

the smaller magnitude represents when the encryption is going on. Should it look like this? I don’t see the distinct spikes for 14 rounds of AES256 encryption. Should not that be visible here? I know there will be some noises but the 14 rounds should be visible/distinguishable i think.

2

It’s really hard to say without knowing the resolution of your X axis and your sampling rate. Can you also show the target trigger line, for reference?

3

I have taken data from oscilloscope and as you have suggested I plotted trigger as well. Below the first image a bit zoomed out. Green line is trigger and red line is probed trace

image
image1920×1457 218 KB

I see several spikes for trigger, 1, 2, 3. I wonder why trigger is behaving like this, spike 2 is confusing me if it really is a proper trigger or not. The next picture is zoomed in

image
image1920×1288 225 KB

I don’t understand if the encryption is going on between 1 and 2 or 1 and 3. whatever the case is, i am failing to see 14 rounds of operations.

4
  1. What is your sampling rate?
  2. What is the target’s clock rate?
  3. What is the trigger and why is it toggling all the time?

Is this a custom AES core, or is it our reference core modified as per this thread?

If it’s a custom core, I suspect that your trigger isn’t behaving as intended; I suggest you run a Verilog simulation to sort that out.

5

@jpthibault I have used the reference core with the settings mentioned in the thread. I was able to get a perfect probing once but can not reproduce it again. As I am not sure what is causing this, I thought I may have done something wrong so checked the AES128 probing using the default setup. The image is attached below:

image
image1920×1243 196 KB

For this image, sampling rate = 50k Sa/s, target’s clock = 10MHz, I have connected the trigger with TP1 as shown in CW305 whitepaper. This is the default AES128 provided by Chipwhisperer. Do you think this is correct trace? Its producing correct encryption.

6

You are sampling 50K times per second, but the target clock is toggling 10M times per second. You can’t expect to see anything useful with such a low sampling rate.

Since you are sampling asynchronously, you should be sampling several times faster than the target clock. 10x should be a good starting point (i.e. 100M Sa/s).

7

@km22, @jpthibault Hello, I am trying to perform a power analysis attack on the CW305 Target, using an Oscilloscope for capturing the traces, I read the provided example , as I will use Oscilloscope so I will not use the scope as the example, I am facing lot of issues could you please help me solve them :

  1. as I am using Oscilloscope, am I still need the capture_traces function to perform the encryption ?, because after specifying the bitstream file for configuring the cw305 to do the encryption, I think I need to make a loop to generate a lot of plaintexts and do the encryption and print the ciphertext right? if so can anyone share the used encrypt_function because in the capture_trace function, they are using scoope.arm() and scoope.get_last_traces .
  2. I didn’t figure out how to plug in the passive probes of my Oscilloscope into the shunt resistor of the boards to capture traces, is there any image I can follow or an example to capture traces with my Oscilloscope?
    Thank you in advance!
8
  1. Explained here.
  2. Best is to use an SMA on X4; you can also use JP6 or JP7, as explained here.