This paper explores the theory and applications of perfectly secure key distribution schemes for dynamic conferences. The authors study a method where a trusted server initially distributes private information to a set of users, enabling any group of a given size (dynamic conference) to compute a common secure key. The key distribution scheme is designed to be secure against coalitions of up to \( k \) users, meaning that even if \( k \) users pool their pieces of information, they cannot compute anything about a key of any \( t \)-size conference. The paper considers two models: non-interactive and interactive. In the non-interactive model, users compute the common key without interaction. The authors prove a lower bound on the size of the user's information, which is \(\binom{k+t-1}{t-1}\) times the size of the common key, and show that this bound is optimal by presenting a scheme that meets it. In the interactive model, users can interact during the common key computation phase, and the authors demonstrate a scheme where the user's information is only \( k+t-1 \) times the size of the common key, showing a gap between the two models. The paper also discusses various applications and modifications of the basic scheme, including hierarchical key distribution, asymmetric user-population, access-control validation, and partial key revocation. Additionally, it presents adaptations to network topologies with neighborhood constraints.This paper explores the theory and applications of perfectly secure key distribution schemes for dynamic conferences. The authors study a method where a trusted server initially distributes private information to a set of users, enabling any group of a given size (dynamic conference) to compute a common secure key. The key distribution scheme is designed to be secure against coalitions of up to \( k \) users, meaning that even if \( k \) users pool their pieces of information, they cannot compute anything about a key of any \( t \)-size conference. The paper considers two models: non-interactive and interactive. In the non-interactive model, users compute the common key without interaction. The authors prove a lower bound on the size of the user's information, which is \(\binom{k+t-1}{t-1}\) times the size of the common key, and show that this bound is optimal by presenting a scheme that meets it. In the interactive model, users can interact during the common key computation phase, and the authors demonstrate a scheme where the user's information is only \( k+t-1 \) times the size of the common key, showing a gap between the two models. The paper also discusses various applications and modifications of the basic scheme, including hierarchical key distribution, asymmetric user-population, access-control validation, and partial key revocation. Additionally, it presents adaptations to network topologies with neighborhood constraints.