Federated Learning (FL) is one of the trending machine learning models that enables multiple entities to process and generate their records in a decentralized environment in an efficient manner. However, it still suffers from several security loopholes that jeopardize the network in seismic manners. Data security is considered as a top priority in today’s era where nearly all crucial records and information are transferred and handled by smart devices over a network. Despite several advancements in capabilities in today’s techniques and frameworks, no organization has shown full confidence in their adoption due to several unresolved security concerns. Though several schemes have been proposed by existing researchers/scientists, however, only few of them have been able to provide security in information transmission while also ensuring accurate prediction of results. The decentralized handling of records in federated learning safeguards each member’s privacy in the network which can be further enhanced in terms of accuracy and security by the proposed scheme. The proposed model integrates an XAI and trust-based model using zero-trust security and SHAP methods to ensure secure and transparent communication among devices in the network. Further, it is verified and validated against several security and performance metrics such as recall, response time, computation and communication time etc.