Abstract
Smart home safety systems increasingly rely on automated alerting mechanisms to detect hazardous conditions such as fires. However, alerts generated by individual IoT devices are often vulnerable to sensor noise, false positives, compromised devices, and limited explainability. This paper presents a zero-trust, quorum-based alert verification framework that integrates lightweight digital twin modeling, decentralized sensor voting, and blockchain-backed auditability to improve the reliability and transparency of fire detection in smart home environments. Each sensor is paired with a digital twin that models expected thermal behavior using a rolling statistical baseline, enabling local deviation-based anomaly detection. Sensor observations are converted into binary votes, and system-wide alerts are validated only when a predefined quorum is reached, preventing any single device from independently triggering an alert. All sensor readings, votes, and final decisions are immutably recorded in an append-only audit ledger. The proposed framework is evaluated using a software-based simulation of a smart home with nine temperature sensors under various environmental setups. Under the evaluated configuration, the system achieves an average verification latency of approximately 2.9 s while maintaining real-time responsiveness and interpretability. These results demonstrate that the framework achieves robust, explainable, and real-time alert verification suitable for smart home safety applications.