[cryptography] Repeated Encryptions Considered.... ?

Sampo Syreeni decoy at iki.fi
Sun Jun 19 18:12:19 EDT 2011

On 2011-06-19, Jack Lloyd wrote:

> There is one case I have seen where encryption with independent 
> ciphers does make sense - for certification reasons. Currently 
> Tahoe-LAFS uses AES to encrypt content, however there is a plan to 
> encrypt all messages first with XSalsa20, then AES, so that side 
> channel attacks on AES are no longer relevant [...]

Do you happen to have a link for the underlying rationale? "No longer 
relevant" sounds a bit strong to me, especially with a cipher as well 
studied as AES is. I mean, wouldn't it be easier to just implement it 
better, and/or to add to the certification requirements?

Now that you gave me the opportunity, I do have to add one point about 
cascaded cipher strength which I forgot to mention. Namely, one of the 
simplest, most common, oldest, and also most fatal mistakes here is that 
symmetric ciphers *can* leak information about the key. Thus, if you 
happen to place a leaky cipher last, it might enable somebody to figure 
out the key, in *particular* if the earlier cipher is strong, so that 
pseudorandomness assumptions apply, statistically speaking. Often you'd 
be using the same key, or the same source data for the key derivation 
function, all over your cascade, which could jeopardize even the 
strongest one in the chain if the last one leaked.

That's an architectural mistake of course, but a common one, and 
something that's rather difficult to avoid e.g. if your key has low 
entropy to begin with. KDFs and especially optimized key schedules are 
rather intricate, after all.

How would you then know which cipher was the strongest, so to be placed 
the last, if you don't know enough to just pick the strongest cipher and 
be done with it without compounding? If you think about the 
probabilities, this possibility not only expectedly undermines the 
cascade. But there's also an argument to be made that those who fudged 
the issue by encrypting multiple times have already signaled that they 
aren't the ones who should be designing the crypto architecture in the 
first place.

This is of course from the book, but I still thought a reminder might 
help the original poster with his question.
Sampo Syreeni, aka decoy - decoy at iki.fi, http://decoy.iki.fi/front
+358-50-5756111, 025E D175 ABE5 027C 9494 EEB0 E090 8BA9 0509 85C2

More information about the cryptography mailing list