Security vulnerabilities in resource-constrained smart-city sensing infrastructures remain a critical barrier to reliable and resilient network operation. In such environments, compromised nodes and insecure routing mechanisms can significantly disrupt data integrity and network stability. Existing approaches often address encryption, trust assessment, or intrusion detection independently, limiting their effectiveness against coordinated and evolving threats. To address these limitations, we propose an integrated security architecture that combines dynamic trust evaluation, machine learning–based node behavior analysis, and blockchain-assisted data integrity. The design incorporates biometric and one-time-password authentication for controlled access, followed by continuous behavioral trust computation. A supervised learning model classifies benign and malicious nodes, while blockchain logging ensures immutable and auditable storage of routing events and transmitted data. Trust scores directly guide routing decisions and enable isolation of compromised nodes. The approach is validated through two heterogeneous simulation scenarios representing traffic monitoring and office environments. Simulation results demonstrate classification accuracy of up to 93% and data integrity exceeding 98% under adversarial conditions. Comparative evaluation against existing state-of-the-art methods confirms competitive detection performance while delivering broader multi-layer security coverage through integrated trust, machine learning, and blockchain mechanisms. Overall, the proposed architecture delivers a scalable and resilient mechanism for enhancing security in smart-city sensor network deployments. The proposed framework is formulated as a computational optimization model integrating trust dynamics and routing cost functions.