Suppose that we have a Bayesian game, where $t_i\in T_i$ denotes the type of player $i$. Say that we have a communication game (communication equilibrium). The players do send each other an encrypted message about their type. If $L_i$ is an isomorphic space of $T_i$ and $\phi_i:T_i\to L_i$ is an permutation (injection + surjection= bijection), then every player $i$ instead of sending their type to each other player they can send the message $\phi_i(t_i)=l_i$. Also, in order to protect herself from cheating, let $\rho_i:L_i\times Y_i\to X_i$ be a cipher, that encodes the private information of player $i$, that is, $y_i\in Y_i$ is the key and $x_i\in X_i$ is the code, where again $\rho_i(\cdot,y_i)$ is a bijective so that the pair $(x_i,y_i)$ is associated with exactly one $l_i$. For technical reasons we made the asumption $|Y_i|\geq|T_i|$ (but why? Is this Shannon's property?).

In our game we have $I$ players, with the above representation we can use a lemma from probability theory, that is:

$\textbf{Lemma:}$ If $\phi_i$ is a random variable with support on $\{1,2,\dots,n_i\}$, and $y_i$ is uniformly distributed over $\{1,2,\dots,n_i\}$ indepedent of $\phi_i$, then the random variable $x_i$ defined as $x_i=\phi_i\ominus_{n_i}y_i$ (where $\phi_i\ominus_{n_i}y_i=\phi_i-y_i(mod{n}_i)$) is also uniformly distributed over $\{1,2,\dots,n_i\}$.

In other words $l_i=\phi_i(t_i)=x_i\oplus_{n_i}y_i$. Then every player $i$ instead of sending $l_i$ to the other agents as a message, she sends in the half of them $x_i$ and the rest of them (we do not know if $I=2k$ or $I=2k+1$, with $k\neq 0$ a potitive integer) $y_i$. Then in a subsequent phase they communicate each one to unite the pieces and verify that they only learn $l_i=\phi_i(t_i)=x_i\oplus_{n_i}y_i$ (however they still haven't learned $t_i$ but $\phi_i(t_i)=l_i$.

My questions are the following

  1. Is this mechanism of information transmission secure? If not, how can I make it?
  2. Could I use a scheme as in secret sharing, where every player $i$ could distribute shares of the key $y_i$ to all of the other players $j\in I-\{i\}$? For example could I further assume that $y_i$ is written as a linear combination of some $w_i$ that all these $w_i$ are non-zero and independent such that $y_i=\sum_{j=1}^{I-1}w_jy_j$? Is this right or wrong? Could anybody provide some help-references-guidance or either show some maths that could make it possible to make such a construction?

In general, how can I enrich this mechanism of communication to become more efficient and secure?


Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Browse other questions tagged or ask your own question.