Since the most critical constituent of the cost of cryptocurrency production is energy bills, the use of illegal electricity in cryptocurrency mining farms is very common. Illegal mining farms have popped up throughout Iran in recent years. They use large collections of computer servers to verify bitcoin transactions, a highly energy-intensive process that can sap hundreds of megawatts from the power grid, which might lead to several large cities facing daily power outages. Therefore, it is essential to detect illegal miners. Although illegal miner detection might seem like a common anomaly detection problem at first glance, the results reported by different power distribution companies in Iran show that the behavior of many normal customers might be very similar to the customers’ that have some illegal miners. In addition, power distribution companies prefer models that can recognize useful insights into the behavioral patterns of the customers. To the best of our knowledge, for the first time, this paper proposes a novel classIfier for miNer detection Based On patteRn miNing (INBORN) that considers the correlation between different attributes and extracts the behavioral patterns of costumers explicitly. INBORN consists of two steps: in the first step, the frequent patterns are extracted and the attributes separating miners and non-miners are determined. In the next step, a decision tree is learned based on the frequency of the patterns. Since the Power Distribution Company of Markazi province is a pioneer in the field of illegal miner detection in Iran, the performance of INBORN is evaluated based on real datasets provided by this company. The experimental results show that INBORN improves the classification accuracy compared to the common algorithms and systems used in the Power Distribution Company of Markazi province.
