Real-Time Traffic Detection for Lightweight Blockchain

Blockchain networks provide trust through decentralization and anonymity but also introduce significant challenges in real-time monitoring and security regulation, with network traffic anomalies being a prevalent threat. To address this challenge, we propose a real-time traffic detection system for lightweight blockchain environments. The proposed system is built on Docker containerization to enable portable node deployment and utilizes Wireshark to gather real-time on-chain traffic data. We extract only three lightweight numerical features i.e., packet index, timestamp, and packet length and address class imbalance using the SMOTE method. For detection modeling, we introduce a hybrid CNN+LSTM+Attention architecture that integrates multi-channel convolutional layers, LSTM, and an attention mechanism to jointly capture spatial-local and temporal dependencies. Evaluated on a dataset of 15,000labeled samples from the Sepolia testnet, our system achieves an accuracy of 84.1 % and an F 1 -score of 0.778. This performance demonstrates the effectiveness and scalability of combining our lightweight feature set with the hybrid deep learning model for containerized real-time detection scenarios.