Fault Tolerance Mechanism for Transient Faults in IoT Based Traffic Data Transaction

The rapid evolution of IoT in smart traffic systems introduces new vulnerabilities, where specifically, transient faults caused by environmental interference and resource constraints. These faults threaten data integrity, system reliability and real-time responsiveness. This paper presents a predictive Fault Tolerance Mechanism (FTM) based on Communication-Induced Checkpointing (CIC) integrated with Long Short-Term Memory (LSTM) networks, tailored for traffic-oriented IoT environments. The proposed Checkpoint at Intermediate Nodes (CIN) CIC-FTM framework places checkpoints at intermediate nodes, based on LSTM-predicted fault likelihood, enabling lightweight and proactive recovery while minimizing rollback overhead. The system architecture designed with IoT edge sensors, fog nodes and a centralized coordination layer to support local fault detection, predictive analytics and consistent checkpoint management. Real-time traffic and communication metadata are used for fault prediction, covering transient faults such as sequence mismatches, checksum failures, presence of null character and out-of-range sensor values. Evaluation across network sizes of 5 to 100 nodes, demonstrates reduced checkpoint frequency by 70-85%, improved fault detection and prediction accuracy by ⁓92% and efficient resource usage. Comparative analysis with existing CIC models confirms significant improvements in recovery time, scalability and adaptability. This hybrid approach combines deep learning, real-time fault detection and selective, proactive checkpointing, offering a robust, energy-efficient and deployment-ready solution for fault tolerant smart traffic infrastructures.