Score:0

Perfect secrecy of the reverse of a crypto system that has perfect secrecy

in flag

I am trying to solve a problem that reads as follows:

Let $E_1 = (\text{Gen}_1, \text{Enc}_1, \text{Dec}_1)$ be a crypto system that has perfect secrecy. Denote the message space $\mathbb M_1$, the key space $\mathbb K_1$ and the cyphertext space $\mathbb C_1$ ($\mathbb M_1=\mathbb C_1 = \mathbb T, \mathbb K_1 = \mathbb K$). Let $E_2 = (\text{Gen}_2, \text{Enc}_2, \text{Dec}_2)$ be a crypto system with the same message space, cyphertext space and key space as in $E_1$, with the only change that $\text{Enc}_2(k,m)=\text{Dec}_1(k,m)$. Does $E_2$ also have perfect secrecy?

What I tried to do is the following:

By definition of perfect secrecy we have that for any distribution $D_1$ over $\mathbb M_1= \mathbb T$ and for any $(m,c)\in \mathbb T \times \mathbb T$ with $\Pr[C_1 = c | M_1 = m]= Pr[C_1 = c]$ where $Pr[M_1 = m] > 0$ ($M_1,C_1$ are random variables).

Now we assume some distribution $D_2$ over $\mathbb M_2=\mathbb T$, and try to prove that this cryptosystem has perfect secrecy like so: Let $m\in \mathbb M_2 = \mathbb T$ such that $Pr[M_2 = m] > 0$, and let $c\in \mathbb C_2 = \mathbb T$, then:

$Pr[C_2 = c| M_2 = m] = Pr[\text{Enc}_2(K,m)=c] = Pr[\text{Dec}_1(K,m)=c] = $

Here there were two ways of going forward. $K$ is a random variable that is distributed by some distribution over $\mathbb K$ and that has to be similar for both crypto systems because they use the same $\text{Gen}$ algorithm. So:

$Pr[\text{Dec}_1(K,m)=c] = \sum_{k:\text{Dec}_1(k,m)=c} Pr[K=k]$ but I don't know what the decoding algorithm looks like so no idea of what these probabilities might be.

Another approach was to assume (perhaps falsely) that $Pr[\text{Dec}_1(K,m)=c] = Pr[M_1 = c | C_1 = m]$ and given the perfect secrecy of $E_1$ that works out to be equal to $Pr[M_1 = c]$

And $Pr[C_2 = c] = \sum_{m\in\mathbb M_2} Pr[M_2 = m]*Pr[C_2 = c | M_2 = m] = \sum_{m\in\mathbb M_2} Pr[M_2 = m]*Pr[M_1 = c] = Pr[M_1 = c]*\sum_{m\in\mathbb M_2} Pr[M_2 = m] = Pr[M_1 = c]*1 = Pr[M_1 = c]$

And so they are equal, therefore $E_2$ has perfect secrecy.

My concern is that the assumption I made before was false, and that $Pr[C_1 = c] could be equal to 0, in which case this is wrong too.

I feel like I'm lacking in understanding what perfect secrecy really means and so I only have this definition in hand to work with..

Any ideas if this is correct?

Thanks in advance.

mangohost

Post an answer

Most people don’t grasp that asking a lot of questions unlocks learning and improves interpersonal bonding. In Alison’s studies, for example, though people could accurately recall how many questions had been asked in their conversations, they didn’t intuit the link between questions and liking. Across four studies, in which participants were engaged in conversations themselves or read transcripts of others’ conversations, people tended not to realize that question asking would influence—or had influenced—the level of amity between the conversationalists.