Privacy-preserving distributed learning has become critical for data-intensive and sensitive domains such as healthcare, smart cities, and industrial IoT. This paper presents a Privacy-Aware Hierarchical Federated Learning (PA-HFL) framework that integrates multi-tier aggregation with Differential Privacy (DP), Secure Multi-Party Computation (SMPC), and Homomorphic Encryption (HE) to address scalability, communication efficiency, and privacy leakage challenges inherent in conventional federated learning systems. The proposed architecture extends traditional two-tier federated learning into a hierarchical structure comprising local clients, edge/regional aggregators, and a global server, enabling decentralized yet coordinated model training while minimizing communication overhead. Privacy-enhancing technologies are strategically deployed across hierarchical layers: DP introduces calibrated noise to protect individual contributions, SMPC ensures secure intermediate computations without revealing private updates, and HE enables encrypted aggregation of model parameters throughout the learning lifecycle. Furthermore, a private blockchain is incorporated to ensure integrity, auditability, and tamper-resistant verification of model updates across all tiers. The framework is evaluated using extensive simulations on benchmark datasets including MNIST, Fashion-MNIST, CIFAR-10, and CIFAR-100, under both IID and non-IID data distributions. Experiments are conducted using Python, PyTorch, TenSEAL, and the Flower federated learning framework on a controlled hardware environment. Performance is assessed using metrics such as model accuracy, communication overhead, computational latency, scalability, and resistance to privacy attacks including reconstruction, membership inference, and property inference. Results demonstrate that the proposed PA-HFL framework achieves higher model accuracy and faster convergence compared to conventional federated averaging, while maintaining strong privacy guarantees.