Thread Links Date Links
Thread Prev Thread Next Thread Index Date Prev Date Next Date Index

RE: towards a formal statement of attack #1 (was RE: is it "randomized"?)



Pardon me Shai: I hope you don't mind me elevating this particular
part of this particular discussion back to the reflector.  I
think it might be useful for others to see this reply.

> Landon Noll wrote:
> > 	An attacker attempts to exploit a flaw in LRW-AES by
> > 	modifying single bits in a block of cipher text in the
> > 	hopes they can flip, with better than pure random
> > 	change, bits is the resulting decryption.
> > 
> > The question is, can such an attack succeed?
> 
> This is an easy one to answer: any success in doing that will 
> be a direct break of AES. From this perspective you should 
> trust LRW-AES at least as much as you do AES itself.

I for one, need to understand how a success on LRW-AES
implies a flaw in AES itself.  Perhaps you are correct,
but at the moment I don't see the connection ...

> When you have a cryptosystem that has a proof of security 
> (such as LRW), the types of attacks that are possible are:
> 
> (a) Attacks outside the model in which the proof was staged, or

Yes.  One of my doubts is if attack #1 is outside the model
in which the proof was staged and ...

> (b) Attacks that succeed with probability (at most) the bounds that
>    were stated in the security proof.

I agree. And there are others:

(c) The scope of the model in which the proof was staged is
    to narrow to be of practical use.

	{sort of a more general case of (a)}

(d) The proof is flawed.

Well hopefully not (d), but then again it wouldn't be the first time
that a flawed proof was published!  :-)

> The attack that you described is clearly covered by the 
> security-proof of LRW. You can read (an upper bound on) the 
> success probability of such attacks directly from the theorem 
> stated in the paper of Liskov et al.

Just to be sure, are you referring to this paper?

	http://www.cs.berkeley.edu/~daw/papers/tweak-crypto02.pdf

If so I'll re-review it in detail over the US holiday weekend.

chongo () /\oo/\