# Correlation power analysis attack on Camellia

I'm investigating a correlation power analysis on a black-box ASIC I have. I want to learn about correlation analysis and it seems this ASIC implements Camellia.

I searched for researches that did CPA on Camellia but it seems no-body did it.

I have 2 questions. 1. why ? is this algorithm resilient to CPA ? if it is then why ? 2. If it does not, which correlation should I search for ? this algorithm is not intuitive like AES ... and also I dont really know that this is camellia. only that this is not AES :)

Also, I assume Camellia since I saw 18 repetitive struct ... maybe 18 rounds ?

Camellia is 128-bit Feistel block cipher and 128/192/256 key. The sbox is 8-bit register, hence there are 8 cells in F function . The correlation power analysis depended on hamming weight. Steps to be taken:

1. Build a matrix of all possible keys (256) on each cell, the matrix
size is 8*256=2048, then calculate the hamming weight ( I guess for
hardware implementation , the model is hamming distance)
2.  Collect power consumption traces during encryption process, the number of traces differ from experiment to other but usually range
up to 10000 traces.
3.   Calculate the correlation between the model and traces.
4.  the best coefficient indicates the possible correct key.
5.  To obtain the 128 key , you need to do the correlation for two rounds (first and second) since the sub key size is 64 bit , the
same applies for 192 (three rounds) and 256 (four rounds)


1.  Is this algorithm resilient to CPA ?

2.   If it does not, which correlation should I search for ? this algorithm is not intuitive like AES .