URL : https://link.springer.com/article/10.1007/s12083-026-02276-w

Title : Integrative framework for driver inattention detection and autonomous safety enhancement leveraging deep learning and blockchain network

Journal : Peer-to-Peer Networking and Applications 

DOI : https://doi.org/10.1007/s12083-026-02290-y

Date : 24 July 2026

Abstract:

This study introduces an integrated system for detecting and managing driver inattention using a combination of artificial intelligence (AI), blockchain technology, and edge computing. The proposed system utilizes You Look Only Once version 8 Nano (YOLOv8n) for real-time object detection and a Raspberry Pi 5 for edge-based analysis, ensuring efficient and accurate detection of inattentive behaviors such as drowsiness, distraction, and other unsafe driving patterns. Blockchain integration with Hyperledger Besu, using the Istanbul Byzantine Fault Tolerance version 2.0 (IBFT 2.0) consensus protocol, provides a secure and tamper-resistant ledger for recording and verifying driver inattention events. The system demonstrated robust performance metrics, including a mean precision of 92.99%, mean recall of 93.47%, and mean Average Precision at Intersection over Union threshold 0.5 (mAP@0.5) of 92.5% for YOLOv8n, establishing its superiority among evaluated YOLO models. The blockchain network achieved a throughput of up to 128.5 transactions per second (TPS) and an average latency ranging from 0.84 seconds for transfer operations to 4.43 seconds for open operations, demonstrating its capability to support real-time event logging. This research addresses limitations in traditional centralized systems by offering a scalable, permissioned, and transparent framework for enhancing driver safety through real-time monitoring and secure data management.