Development of machine learning methods and algorithms for detecting terrorist financing activities in the Republic of Kazakhstan

Today, the terrorist activity of organizations, groups and individuals is recognized as one of the main sources of threats to the national security of the Republic of Kazakhstan. The high level of this threat is primarily due to the scale of the consequences of terrorist crimes and the significant number of victims. It suffices to recall the recent terrorist attacks in Aktobe and Taraz and the events of January 2022 in Almaty. The intensity of terrorist activity depends directly on the level of its financing and material and technical resources. Therefore, freezing the assets of terrorist organizations and blocking terrorist financing channels are recognized as among the most important tools in the fight against international terrorism. The systems for detecting suspicious (anomalous) behavior currently used by financial institutions are mainly based on sets of rules defined by experts. Such rules are neither flexible nor easily adaptable, which makes them vulnerable to abuse and circumvention. Another problem is that these systems produce a large number of false positives, which take time to process. Machine learning-based solutions can be continuously trained and updated with new data, which makes them adaptable and eliminates the need for experts to define new rules.

Project Objectives

1. Understanding the dataset: exploring the dataset and each of its variables with the necessary visualizations.
2. Exploratory data analysis: conducting the analysis and drawing critical conclusions with a full description of the data slices considered; formulating hypotheses.
3. Visualizations and charts showing the difference between fraudulent and non-fraudulent transactions.
4. To develop methods of automatic annotation and interpretation that make it possible to identify the largest possible set of transaction anomalies.
5. Modeling: testing the methods, the iterations used in the experiments and the data transformations.
6. Development of a system to provide higher authorities with recommendations on diagnostics for further work with anomalous transactions.
7. Testing the system on synthetic data obtained from platforms such as AMLSim.

The project is subject to further commercialization.