<p>Secure water quality monitoring is a critical requirement for smart urban infrastructures. It is increasingly critical to ensure timely contamination detection and trustworthy data management for public health. However, existing IoT-based monitoring systems often rely on centralized architectures, making them vulnerable to data manipulation. Such systems are also afflicted with scalability limitations, and delayed response to contamination events. This study proposes a unified framework that integrates IoT sensing, blockchain technology, smart contracts, and a hybrid deep learning model for secure and real-time water quality monitoring. The proposed system employs a Multilayer Perceptron–Gated Recurrent Unit (MLP–GRU) model to capture nonlinear relationships and temporal dependencies in physicochemical parameters. To ensure data integrity and automated validation, smart contracts are deployed on a blockchain network. Simultaneously, large-scale sensor data are managed using off-chain storage through IPFS. Squirrel Search Optimization (SSO) is applied to optimize model hyperparameters, improving convergence stability and predictive performance. Experimental evaluation demonstrates that the proposed framework achieves strong predictive performance, with an accuracy of 98.17% ± 0.15%, precision of 97.00% ± 0.22%, recall of 97.48% ± 0.18%, and F1-score of 97.32% ± 0.20%. These results show that the proposed system outperforms conventional models including LSTM-based and federated learning approaches. Furthermore, the system achieves reduced latency (87 ms) and a 95.2% reduction in false alarm rates, highlighting its effectiveness for real-time monitoring scenarios. The integration of predictive intelligence with blockchain-based validation enables a secure, transparent, and scalable monitoring solution. The proposed framework provides a practical approach for reliable water quality management, supporting automated decision-making and enhanced trust in next-generation smart city environments.</p>