I'm looking for course material on the subject of proofs, reductions, and games, as used to prove cryptographic schemes secure. What are the methodologies? What are the preferred ones? In what cases is it best to use a method rather than another?

  • 2
    $\begingroup$ Literature recommendations and reference requests are off topic here. I'll leave it open for now, but please edit to make more on topic. That said, I think there is a very good question that could be asked here. Instead of asking for references, I'd suggest editing the question to ask what the various proof techniques are, their strengths/weaknesses, etc. In that I think you will get lots of good references plus some good initial information to help you on your journey. $\endgroup$
    – mikeazo
    Jun 7, 2013 at 13:53

1 Answer 1


The two primary techniques I'm familiar with is structuring a cryptographic primitive as a sequence of games and the universally composable security framework.

Sequence of Games
The idea here is to represent a protocol/primitive as a game played between an attacker and a challenger. You define a bad event and show through the game that the event happens with (close to) some target probability (say 1/2 or 0). For example, if we are modeling an RNG and want to show that the attacker can only guess the next output bit with probability very close to (formally, negligibly close to) 1/2. If they are able to do better than that, then the RNG isn't very good. 1/2 is chosen since there are only 2 choices for the next bit (0 or 1) and they should occur each with probability approximately 1/2 so an attacker who simply always guesses 0, for example, should be right 50% of the time.

The paper I linked to above has some good examples.

UC Framework
This is a framework proposed by Canetti that sees a lot of use in protocols, especially multiparty computation, but is useful in many situations. The UC framework defines a scenario where we have a real world with an environment $\mathcal{Z}$, an adversary $\mathcal{A}$ and an ideal world with the same environment, a simulator $\mathcal{S}$ and an ideal functionality $\mathcal{F}$. In each world, we also have the participant parties $p_1,\dots,p_n$. The setup is pictured below (taken from Martin Geisler's dissertation): UC Setup

In the real world, the parties execute some protocol $\pi$, in the ideal world the ideal functionality $\mathcal{F}$ provides the same functionality as $\pi$ but is secure by definition (think a trusted, uncorruptable third party). The UC Security theorem says that $\pi$ securely realizes $\mathcal{F}$ in the real world if there exists a simulator $\mathcal{S}$ in the ideal world such that $\mathcal{Z}$ cannot distingush the two worlds. In other words, if $\mathcal{Z}$ cannot tell which world they are operating in, then $\pi$ must be at least as secure as $\mathcal{F}$. But $\mathcal{F}$ is as secure as possible by definition, so in this case $\pi$ is also secure.

Another cool thing about the UC framework is the composability theorem which says that UC-secure protocols can be composed with other UC-secure protocols (including itself) in arbitrary (even adversary controlled) ways and the resulting composition is also UC-secure. Canetti's paper above has examples.

A third technique that is often used is a reduction proof. This is very similar to reduction proofs in theoretical computer science. Basically it says we have some problem $X$ which is hard to break (how exactly that is determined is beyond the scope of the answer, but usually it is that a bunch of really smart people have been trying to break it and haven't). Say I have a new cryptosystem and have shown that breaking my cryptosystem is equivalent to breaking $Y$, i.e., if you break $Y$ you can break my problem. Then, if I can show that breaking $Y$ is at least as hard as breaking $X$, then there is reason to believe my cryptosystem is also secure. For example, see this answer.

  • $\begingroup$ @user7060, do you plan on updating the question to address the concerns I raised? If not, let me know and maybe I'll update it. $\endgroup$
    – mikeazo
    Jun 8, 2013 at 1:11
  • 1
    $\begingroup$ The UC framework is an instance of the "simulation-based" paradigm for security definitions. This paradigm first appeared in Goldreich, Micali & Wigderson (STOC 87), building on work of Goldwasser, Micali & Rackoff (STOC 85). The general idea is that, in contrast to defining a security game with one adversary in terms of a success probability, one defines a "real" interaction and asks that for every adversary there exist a simulator, operating in an "ideal" interaction that can achieve the same effect. Not all simulation-paradigm definitions have the composability properties of UC. $\endgroup$
    – Mikero
    Jun 10, 2013 at 5:32
  • $\begingroup$ There are no cryptographic primitives that have proof of security unless you can show that P is not equal to NP. Just showing a reduction from some cryptographic primitive to a known hard problem in NP does not mean that the primitive is secure, you also have to prove that the problem in NP is computationally intractable. That means separating P from NP. This is well known in theoretical computer science. $\endgroup$ Jun 20, 2013 at 4:38
  • $\begingroup$ @mikeazo I don't agree very much with the categorization as the sequence of games are part of a more general category which is the indistinguishability game played between the attacker and the simulator-oracle (IND-CPA,IND-CCA,IND-CCA,) $\endgroup$
    – curious
    Jan 30, 2014 at 11:02

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Not the answer you're looking for? Browse other questions tagged or ask your own question.