Mutual learning-based group synchronization of neural networks

In this study, a method for securing neural key exchange over public channels by synchronizing a set of neural networks is presented. To share the key through a public network, two artificial neural networks (ANNs) are coordinated via mutual learning. The most crucial component of neural cooperation is establishing how effectively two parties’ ANNs coordinate without the other’s weights. Furthermore, research on the mutual training of a cluster of ANNs is limited. This research proposes a collaborative learning technique for measuring the perfect coordination of a collection of ANNs quickly and efficiently. Collaboration is determined by the frequency by which the two networks have had identical results in prior sessions. A full binary tree framework is required to organize the ANNs. In the tree architecture, each ANN is a component. A leaf node is a component that has no successors. This technique has numerous benefits such as (1) the session key is generated via full binary tree-based group mutual neural coordination of ANNs over the public channel. (2) The presented scheme, in contrast to prior methods, allows two communicating entities to identify complete coordination more rapidly. (3) In addition, the suggested approach allows for simultaneous coordination and authentication. The adversary has a hard time distinguishing between coordination and authentication stages. As a result, the adversary has no idea yet if the presently visible output bit is utilized for coordination or authentication. (4) The group’s coordinated shared weights act as a key for the session for the whole group to exchange data. (5) The proposed approach takes brute force, geometry, and majority attacks into account. The performance of the suggested methodology is tested, and the results show that it outperforms similar strategies already in use.