Exploring Anomalous Transactions without Disclosing Data

The origin of this experiment was from a visit to MIT Media Lab, which had the privilege of hearing inspiring and creative notions from exceptionally smart people, with few that left a lasting effect. The idea of “Sharing insight without sharing data” and “Wisdom on Demand” were two such concepts. This experiment was based on the former, while ChatGPT’s post-singularity was affiliated with the latter. The FinTech implemented here managed to secure deals due to its transaction security, and was later taken over by a larger technology company. This experiment seeks to determine if there are ways to enhance Anti-Money Laundering (AML)/Counter Terrorism Financing (CTF) transaction monitoring capabilities through the use of innovative data analysis techniques, which could allow for sharing of insights about datasets while still maintaining the secrecy of the dataset itself. Transaction monitoring with AML/CTF capabilities aims to combat the utilization of both domestic and international financial systems for unwholesome economic transactions such as drug money laundering, illegal gaming, exploitation of the vulnerable, cybercrime, and many others. It also looks to deter the routing of funds to support terrorism, as declared by influential Western nations. The severity of penalties for supporting or encouraging these unacceptable activities can range from increased capital requirements, to regulators issuing warnings, to regulatory punishments, to fines that equate to multiple year's worth of income and even being denied access to prestigious resources such as a reserve currency and the SWIFT messaging system. Hence, most financial institutions expend a great amount of resources towards strengthening their Anti-Money Laundering/Counter-Terrorist Financing ability within their transaction surveillance division, in order both to detect and impede forbidden financial practices, and to show that they were conscientious in gaining the capacity to do so. However, FIs face an acute challenge when it comes to detecting AML/CTF activities. This is because the agents engaged in such activities do not use only one financial institution for their entire transaction set. Instead, they break the set down into smaller and more granular subsets and route them through various FIs. This may involve forming pathways in advance, randomly navigating obstacles, or introducing other methods to obscure the intent. Even some legitimate transactions for economic activities may be involved. From the FI's perspective, there could be an issue when they have a view of a subset of transactions which are innocuous when viewed on their own, however, become malignant when combined with other subsets. Even when suspicious activities are suspected, FIs are unable to talk to other FIs in the transaction chain to verify or dismiss the suspicions. Banking secrecy prevents this communication, as FIs have a responsibility to keep their customers' data confidential. Therefore, they cannot openly discuss and share transaction data in an attempt to get a full picture of the situation. The FinTech was of the opinion that novel data analytics could be used to close the gap between FIs by allowing them to exchange insights into transactions while still adhering to their confidentiality obligations. The basic principle is that each set of transactions can be condensed into a transaction pattern, a summary representation of how the transactions occur. Privacy is preserved through the utilization of pseudo-anonymous information about the entities involved, which serves to identify those partaking in the transaction pattern. For example, in the following transaction subset consisting of (1) Adam transferring S$200 to Billy; (2) Charlie passing along S$200 to Billy; (3) Billy sending S$300 to Donald; (4) Donald transmitting S$200 to Charlie; (5) Donald sending S$100 to Adam, the true intent is to cause the outcome that Adam is left S$100 poorer and Billy S$100 richer, with additional dealings being used to obscure the genuine purpose. The following transaction cycle will be carried out: (1) A transferring S$200 to B; (2) C giving S$200 to B; (3) B transferring S$300 to D; (4) D passing S$200 on to C; (5) D transferring S$100 to A. This kind of pseudo-anonymous transaction pattern can be provided to other financial institutions, to verify whether or not any similar patterns of these transfers can be found in their set of transactions. Matching transaction sets could be an indication of the ulterior motives behind these transfers. The FinTech sought to reach this objective by building a transaction pattern database, where FIs can upload a segment of dubious transactions. If the experiment is successful, it can create a chance for FIs to share information about suspicious transactions without exposing the details of those transactions. When combined with data analytics, this would enable the FIs to build up a bigger picture of a set of transactions, and potentially decide whether any further action is needed.

Source: https://medium.com/@notherArmchairEconomist/experiment-to-share-insights-about-anomalous-transactions-without-sharing-data-4921d9e2d56c?source=rss------fintech-5