Question about your power analysis hardware setup (SecTalks talk)

[inou44]

Hi Norm,
I hope you don’t mind me reaching out here, I couldn’t find your email. I came across your SecTalks presentation on side-channel attacks (https://www.youtube.com/watch?v=apm03qHNuG4&t=888s), and I was curious about your hardware setup.

I’m currently working on a small project where I’m trying to perform a Simple Power Analysis on a naïve RSA implementation, with the goal of distinguishing square vs multiply operations to recover exponent bits. My setup is inspired by this tuto: https://github.com/lord-feistel/power_analysis.

I initially tried measuring power on an Arduino Uno using a shunt resistor and an oscilloscope, but the results were too noisy. I then moved to a bare ATmega328P on a breadboard to get a cleaner signal, but I’m still struggling to clearly observe variations. I suspect it could be due to my measurement setup (my shunt value was 1ohm).

I had a few quick questions about your setup:

• What shunt resistor value did you use?
• Did you keep decoupling capacitors near VCC, or reduce them for measurement?
• Were you measuring on the high side or low side?
• Did you use any amplification or filtering before the oscilloscope?
• Do you think soldering the shunt directly on the Atmega makes a significant difference compared to a breadboard setup?

Any guidance would really help, I’m trying to understand what matters most to get a visible signal. I’d really appreciate any advice on how to obtain a clear current consumption trace for such small variations.

Thanks a lot for your time, and for sharing your work!

[CreateRemoteThread]

Hey, did you end up getting this worked out?
If it helps:

• I think from memory for atmega328 I used something like a 50-100 ohm resistor
• I didn't keep (well, didn't use) decoupling capacitors because the chip worked fine without them
• I can't remember high side vs low, don't think it made a difference. low side i think is easier, because you can just measure from 0 to resistor (vs having to subtract vcc from a high side measurement)
• I didn't amplify or filter
• I don't think soldering the shunt directly makes a difference

If you can't see the signal at all, and your measurement setup is otherwise ok (i.e. using a trigger signal to begin measuring, voltage and time scale ok - should be a few mV with just a low side shunt?), try increasing your shunt resistor value, sometimes that helps