Analyzing Large-Scale Blockchain Transaction Graphs for Fraudulent Activities

Abstract

Abstract Early public blockchains provided low transaction throughputs in the range of 7–30 transactions per second. With the emergence of permissioned and proof-of-stake-based blockchains, transaction throughputs are expected to rise drastically to thousands per second. Blockchain transactions form directed graphs. With high transaction throughputs and growing blockchain adoption by banks, businesses and customers in general, the number of edges in transaction graphs will dynamically grow to billions. An analysis of large-scale transaction graphs is needed for tracing fraudulent activities on blockchains. This chapter will cover topics such as distributed graph data structures, the use of message passing libraries, and parallel graph algorithms in order to build a scalable transaction graph analysis system. Results from the analysis of the real Ethereum and Bitcoin public blockchain data involving cryptocurrency and ERC20 token transactions will be presented.