Trustworthy and explainable federated system for extracting descriptive rules in a data environment

A connected world in an information age with dozens of connected devices per person constantly generates continuous data streams. This leads us to the need to generate new intelligent models that discover knowledge in complex paradigms. However, these complex paradigms (capable of generating knowledge in isolated devices and sharing it between them, known as federated learning) must comply with the guidelines of Trustworthy Artificial Intelligence that obtains models with high levels of security, privacy, explainability and traceability. This contribution introduces the Trustworthy and Explainable Federated System based on Supervised Descriptive Rules for Data Streaming (TEFeS-SDR) algorithm, a trustworthy and explainable federated system for extracting descriptive rules in streaming data environments. This model, based on federated learning, emphasizes privacy and security through binary encoding and asymmetric encryption, avoiding the transfer of raw data between devices. Additionally, the system ensures traceability and auditability of the generated rules, providing transparency and trust. Experimental results demonstrate its capability to handle abrupt changes in data streams (concept drift) while maintaining high-quality and homogeneous global models. This work advances the path towards responsible artificial intelligence by combining explainability, security, and efficiency in dynamic environments.