An Adaptive Data Security Frame Using Federated Learning and Blockchain for Privacy Protection in Smart Environments

Abstract

The rapid expansion of smart environments, including smart cities, healthcare systems, and intelligent energy grids, has resulted in the generation of massive volumes of distributed and privacy-sensitive data. Conventional centralized security architectures are increasingly inadequate to guarantee confidentiality, integrity, and trust under adversarial and resource-constrained conditions. This paper proposes an Adaptive Federated Blockchain Security Framework (AFBSF) that integrates federated learning (FL) with a lightweight blockchain layer and a dynamic trust-driven cryptographic control mechanism. Federated learning enables collaborative model training without sharing raw data, while a Proof-of-Authority (PoA) blockchain provides tamper-resistant verification and transparent auditability of model updates. In addition, an adaptive trust model dynamically adjusts encryption strength and node participation according to behavioral reliability and data integrity, allowing real-time isolation of malicious or unreliable devices. Extensive experiments conducted on smart healthcare, energy, and transportation datasets demonstrate that the proposed framework outperforms conventional FL-based, blockchain-based, and existing hybrid approaches in terms of accuracy, privacy preservation, communication efficiency, and energy consumption. The results confirm that AFBSF achieves high learning performance with enhanced privacy protection, reduced attack success rate, and lower system overhead, making it a scalable and reliable security paradigm for next-generation decentralized IoT ecosystems.