Improving cryptocurrency money laundering detection through reinforcement learning techniques

Abstract

The use of bitcoin for money laundering, as is the case with the traditional world, ranks among the most essential threats to the integrity of the banking system, given that the number of channels and gates reside in the decentralized space of bitcoin. In the earlier days when basic recognition systems were deployed, their working principle hinged on the use of massive pre-labeled datasets. But at the moment, the case is somewhat different, as the average bitcoin wallet usage comes from anonymous users. This paper proposes a new exploratory framework for the detection of laundering activities in bitcoin exchanges based on (RL) reinforcement learning. This method however, helps in avoiding the problem of data scarcity by identifying models from both labeled as well as unlabeled transactions and in finding any patterns indicative of illicit act.