Financial Fraud Detection: A Declarative Approach
Name
Mohga Soliman Emam Hewashy
Abstract
The aim of the thesis is to introduce a declarative approach for the statically
specified financial fraud detection use cases and scenarios defined by the financial regulatory entities to capture money laundering and terrorist financing activities ML/TF. The thesis introduces the Match_Recognize Python library that replaces the static rules ingested into the financial institutions and organizations transaction monitoring system to detect financial fraudulence and suspicious activities. The introduced Match_Recognize Python library mimics the functionality of the SQL Match_Recognize clause which performs pattern recognition using regular expressions which can be used to detect financial fraud patterns and therefore eliminate the need to design and develop dedicated static use case scenarios. Using the Match_Recognize library, financial institutions and organizations can produce the financial fraud detection use case scenarios required by the financial regulatory entities using simple regular expressions that are passed to the library alongside the dataset. Additionally, the Match_Recognize Python library contains a Match_Recognize Automaton function
that validates new, proposed patterns in the form of regular expressions within the Match_Recognize clause pattern regular expression by using the non-deterministic Automaton created dynamically from the Match_Recognize pattern regular expression.
The thesis also introduces versatile, pliant financial fraud detection scenarios inspired by the Financial Action Task Force Recommendations. Evaluation of the
Match_Recognize Python library is conducted by running the financial fraud detection scenarios on both the Match_Recognize clause in Oracle database and
Match_Recognize Python Library then comparing the results. A dedicated time log has been created in order to compare the averaged time taken to simulate each financial fraud detection scenario statically and to simulate it using the Match_Recognize Python library. The results show that indeed the Match_Recognize library reduces the financial fraud scenario simulations time by 96.3%.
specified financial fraud detection use cases and scenarios defined by the financial regulatory entities to capture money laundering and terrorist financing activities ML/TF. The thesis introduces the Match_Recognize Python library that replaces the static rules ingested into the financial institutions and organizations transaction monitoring system to detect financial fraudulence and suspicious activities. The introduced Match_Recognize Python library mimics the functionality of the SQL Match_Recognize clause which performs pattern recognition using regular expressions which can be used to detect financial fraud patterns and therefore eliminate the need to design and develop dedicated static use case scenarios. Using the Match_Recognize library, financial institutions and organizations can produce the financial fraud detection use case scenarios required by the financial regulatory entities using simple regular expressions that are passed to the library alongside the dataset. Additionally, the Match_Recognize Python library contains a Match_Recognize Automaton function
that validates new, proposed patterns in the form of regular expressions within the Match_Recognize clause pattern regular expression by using the non-deterministic Automaton created dynamically from the Match_Recognize pattern regular expression.
The thesis also introduces versatile, pliant financial fraud detection scenarios inspired by the Financial Action Task Force Recommendations. Evaluation of the
Match_Recognize Python library is conducted by running the financial fraud detection scenarios on both the Match_Recognize clause in Oracle database and
Match_Recognize Python Library then comparing the results. A dedicated time log has been created in order to compare the averaged time taken to simulate each financial fraud detection scenario statically and to simulate it using the Match_Recognize Python library. The results show that indeed the Match_Recognize library reduces the financial fraud scenario simulations time by 96.3%.
Graduation Thesis language
English
Graduation Thesis type
Master - Computer Science
Supervisor(s)
Ahmed Awad
Defence year
2023