The government?s effort to share financial transaction data among its various law enforcement and intelligence agencies is a necessary step in the battle against organized crime and terrorism. However, the reliance on automated data mining of a single source of data must not be viewed as a silver bullet.
Certainly, there are advantages to data mining the vast amount of financial transactions flowing in and out of the United States ; however, it must be understood that they are limitations to this effort. According to security expert Bruce Schneier , ?data mining works best when you’re searching for a well-defined profile, a reasonable number of attacks per year and a low cost of false alarms.?
Accordingly, the CEO of the company that provides analytics software to FinCEN, Christopher Westphal , says, ?Even if we come up with the perfect rule to expose the perfect pattern, [money launderers and terrorists] are going to change, and we have to be adaptive enough to recognize that and be able to act on it.? Simply stated, the profile of a suspicious financial transaction must evolve constantly because terrorist and organized criminals will change their tactics to avoid investigation by law enforcement and intelligence agencies. A Treasury Department report, entitled US Money Laundering Threat Assessment, highlights this point dramatically. Specifically, the report states: ?the volume of dirty money circulating through the United States is undeniably vast, and criminals are enjoying new advantages with globalization and the advent of new financial services such as stored-value cards and online payment systems.? Therefore, analysts must be on the lookout for new patterns in money transfer such as the use of stored-value cards and online payment systems. An overreliance on data mining software may miss these changes in tactics unless analysts update the profile of what constitutes a suspicious transaction.
In addition to being flexible enough to adapt to changing tactics, the algorithm used to detect suspicious financial transaction must also be specific enough to weed through the increased volume of financial transaction data. A number of factors will contribute to a marked rise in the volume of financial transaction data flowing through these systems in the short-term. First, more stringent reporting requirements that demand untraditional financial institutions, such as casinos, to report financial transactions will increase the volume of data available to analysts. Second, the integration of multiple financial transactions data sources across various government organizations will further increase the mass of data for analysts to comb through. Therefore, there will be an exponential rise in the volume of data as a result of the increased reporting requirements and integration of multiple financial transaction data sources. Given the massive amount of data that must be analyzed, any data mining algorithm must be finely tuned so that an excessive amount of false positives are not generated. An excessive amount of false positives would more than likely overwhelm the analysts and cause them to miss the one suspicious transaction hiding among the thousands of false flags.
Given these realities, data mining the trove of financial transactions will only work in combination with data from other intelligence sources. A hit from the financial transaction database should be measured against information from the State Department, CIA, NSA, FBI, DHS, and other government agencies. For example, a suspicious financial transaction should be compared against NSA surveillance data, FBI reports, and foreign intelligence from CIA liaisons. This cross-reference of sources will help reduce the amount of false positives and negatives. Therefore, the concerns about tuning the data mining algorithm can be somewhat alleviated, as a number of false positives can be eliminated by cross-referencing other sources of non-financial transaction data.