AI-Enabled Abnormal Behavior Detection and Visualization Technology on Blockchain Network

Blockchain is a technology that ensures integrity, transparency, and reliability of transactions without a centralized server by jointly verifying, recording, and storing transaction information by all participants in the network. As the value of blockchain increases, various security threats are emerging in blockchain network. In addition, it is important to research on detecting anomaly behavior in the blockchain network in a short time. In this paper, in order to detect blockchain network security threats and abnormal behavior detection elements, we collect data based on transaction information and present experimental results for anomaly detection using unsupervised learning. The learning model performed a simulation to determine normal or abnormal behavior for about 17,000 blocks per second with an anomaly detection time of 0.058ms per block. The performance of the learning model is an accuracy of 98.2%, false negative 0.6%, and false positive 1.2%, and it correctly detects 99.4 blocks out of 100 abnormal behavior blocks as abnormal behavior. In addition, 99.8 blocks out of about 100 blocks generated by normal behavior are detected as normal behavior. In total, when 100 blocks were generated, normal or abnormal behavior was correctly detected for 98 blocks.