CPA attack on UFO STM32F2

Hi,

I’ve been running some AES CPA tests on various UFO boards, in particular the XMEGA and the STM32F2 (code based on the PA_CPA_1 tutorial). The STM32F2 has a HW AES engine, that’s why I bought it (https://no.mouser.com/ProductDetail/NewAE/NAE-CW308T-STM32F2HWC?qs=sGAEpiMZZMu3sxpa5v1qrvyz3TtLhlhtenSqsOE1ilk%3D). When I run the CPA attack on the XMEGA board I need around 40 traces to recover all subkeys when using an SMA connection, and around 1500 traces when using the H probe (placed center and right on top of IC). For the STM32F2 board, however, I only need around 12 traces for the SMA connection, and 34 000 traces with the H probe. I can understand that I’ll be needing far more traces to recover all the subkeys for the STM32F2 compared to the XMEGA when using the H probe (because of the HW AES), but why is it the opposite when using the SMA connection? Is there a logical explanation for that?

Maybe you got “lucky” with the probe placement, or with the random values that got picked for the attack?

Jean-Pierre