I'm working on side channel attack on prime number generation by measuring power consumed by the cpu while doing that, let's say that the cpu process 10000 mpz_t integer ( size 512 ) does the number of set bits (Hamming weight) in those integers have anything to do with power consumption by the cpu?, in other words a 512 bits integer with 500 set bits is going to consume more power than a 512 bits integer with only 10 set bits?
Asked
Active
Viewed 75 times
1
-
2Isn't that the exact question your experiment should be designed to answer? – zwol Jan 29 '17 at 17:45
-
I don't think, by the way, this is a not a very useful question. – Stargateur Jan 29 '17 at 17:52
-
yes it is, all i have now is a graphe of power consumed by the cpu, and i'm suppossed to extarct enough information, if set bits have some sort of correlation with power consumption i'll be able to find the prime generated, – Hamza Tahiri Jan 29 '17 at 18:04
1 Answers
0
Yes, that's the general idea. There are of course tricks to minimize the difference between a 0 and 1 bit, but without measurements against side channel attacks, a zero bit will be faster than a bit set to one during (modular) multiplication - basically because, for multiplication - nothing needs to be calculated.
Maarten Bodewes
- 90,524
- 13
- 150
- 263
-
exactly because the professor didnt say anything about how primes will be calculated, but in optimized modular multiplication ( form the left to the right) nothing need to be calculated if a bit is a zero, but that's means the attack only works if the methode used modular multilication, thank you for your answer. – Hamza Tahiri Jan 29 '17 at 18:36
-
i'm wondring if this attack could work with diffrent type of prime generation? – Hamza Tahiri Jan 29 '17 at 18:41
-
Any prime generation will have to test for primality of the generated number, and that will always, in some form, require operations such as multiplication and division on the number. I don't see how you could avoid that. – Maarten Bodewes Jan 29 '17 at 20:33