<p>The convergence of sixth generation (6G) networks and the Internet of Things (IoT) is expected to revolutionize smart home environments through ultra-low latency, large-scale connectivity, and intelligent automation. However, these advancements raise increasing concerns about data privacy and security—stemming from the massive volume of sensitive behavioral data generated by interconnected devices, vulnerability to cyberattacks, unauthorized access, and centralized points of failure—as well as reliable behavioral monitoring across various devices. Existing approaches based on blockchain, federated learning, and edge computing have attempted to address these challenges; however, they remain limited by scalability constraints, high computational overhead, integration complexity, and inadequate temporal modeling for behavior prediction. To bridge these gaps, this study presents a privacy-preserving 6G-IoT framework that integrates deep learning with blockchain technology. A hybrid Autoencoder–Temporal Fusion Transformer (TFT) model is employed to learn user behavior patterns, predict activities, and detect anomalies, while the licensed Hyperledger Fabric blockchain ensures secure, tamper-proof, and decentralized data management. The framework includes an identity-based access control system, immutable event logging, and encrypted off-chain storage to enhance privacy and integrity. Experimental evaluations demonstrated the superiority of the proposed system over existing methods, achieving 98.57% accuracy, 97.89% recall, 99.15% precision, and 98.51% F1-Score. Accuracy, recall, and Receiver Operating Characteristic (ROC) curve analyses further confirmed its robustness, with an average area under the curve (AUC) of 0.998. Overall, this work presents a scalable, secure, and high-performance solution for intelligent behavior analysis and anomaly detection in next-generation 6G-enabled smart homes.</p>