The rapid evolution of financial technology demands constantly improving methods for detecting fraudulent transactions and money laundering schemes. This paper introduces a graph-based system utilizing expert rules for evaluating financial transactions, addressing the gap between complex automated solutions and practical implementation needs. The research involving 55,778 financial transactions, analyzed via ArangoDB and AQL queries, establishes that expert rules effectively identify suspicious activities through transaction attributes, frequency patterns, and account characteristics. The results show that different rule configurations identified potentially fraudulent transactions at rates ranging from 0.0072% to 1.38% of the dataset. The study also identifies several limitations, including the need for regular rule updates to address evolving fraud patterns and the system’s dependence on predefined expert knowledge. Nevertheless, the approach demonstrates practical value as an efficient preliminary screening tool that can be implemented by financial institutions and regulatory bodies to flag suspicious transactions for further investigation.