Let $\varepsilon>0$ be a constant. Say an encryption scheme is $\varepsilon$-perfectly secret if for every adversary $\mathcal{A}$ it holds that $$ \operatorname{Pr}\left[\operatorname{PrivK}_{\mathcal{A}, \Pi}^{\mathrm{eav}}=1\right] \leq \frac{1}{2}+\varepsilon $$ Consider a variant of the one-time pad where $\mathcal{M}=\{0,1\}^{\ell}$ and the key is chosen uniformly from an arbitrary set $\mathcal{K} \subseteq\{0,1\}^{\ell}$ with $|\mathcal{K}|=(1-\varepsilon) \cdot 2^{\ell} ;$ encryption and decryption are otherwise the same. (a) Prove that this scheme is $\varepsilon$-perfectly secret. (b) Prove that this scheme is $\left(\frac{\varepsilon}{2(1-\varepsilon)}\right)$-perfectly secret when $\varepsilon \leq 1 / 2$ (c) Prove that any deterministic scheme that is $\varepsilon$-perfectly secret must have $|\mathcal{K}| \geq(1-2 \varepsilon) \cdot|\mathcal{M}| $
This is an exercise from Introduction to Modern Cryptography that I'm studying I already found Proving that a scheme is $\epsilon$-perfectly secret but I need to understand this in detail, Is there anyone who can explain it to me?
