Cost-Effective Anomaly Detection for Blockchain Transactions Using Unsupervised Learning

If the user performs an online transaction, it is disseminated to the blockchain network (BN) and irrevocably implemented. But if a malicious user signs a fake deal, it cannot be half-done far along, thus causing a loss of resources for the customer. Hence, to reduce the processing power and bandwidth, we propose to design a cost-effective anomaly detection for blockchain transactions using unsupervised learning techniques in this paper. In this technique, the nodes in the BN are grouped into various clusters. If any node of a cluster has missed transactions, it can be reconstructed using the erasure coding technique. Then the transaction history is grouped based on the address. The time series data features are extracted by applying the rolling window aggregation to detect abnormal transactions. Then, K-means clustering is used to perform anomaly detection. By simulation results, we show that the proposed approach is cost-effective