The lack of labeled data in the new generation of blockchain networks, such as XRP Ledger (XRPL), which also suffers from the lack of labeled data, is a notable deficiency. This work addresses the well-known “labeled data bottleneck problem” using Cross-Network Transfer Learning, postulating the universality of the fraud network structure. This research investigates the efficiency of three different adaptation mechanisms, Renaming, Scaled Transformation, and Synthetic Generation, to successfully transfer data from Bitcoin network to XRPL network. Experimental results indicate the attainment of 0.927 and 0.878 F1-score performances for transactions and wallet classification models, respectively when they are trained from transformed data. The solution also achieves the best performance using real XRPL distribution parameters with the Distribution-Based Synthetic Generation approach, which confirms the feasibility of the proposed approach.