Skip to main content

Federated Learning In Banking

Money laundering poses a significant societal threat as it enables criminals to utilize illicit funds, undermines public trust, and damages the financial system. To combat money laundering, collaboration between banks is imperative but is currently hindered due to the sensitive nature of transactional data. This project aims to unlock the potential of collaboration between banks by leveraging the recent machine learning paradigm of federated learning.

Federated Learning In Banking


The financial industry confronts multiple challenges in effectively detecting and preventing money laundering while adhering to EU anti-money laundering directives. Some of these challenges encompass:

  1. Handling sensitive data: Guaranteeing the privacy and security of sensitive transaction data while facilitating efficient data analysis.
  2. Facilitating cooperation between banks: Surmounting legal and technical barriers to enable collaboration and information sharing.
  3. Adapting to evolving criminal strategies: Staying ahead of sophisticated money laundering schemes that continuously adapt to exploit emerging vulnerabilities.
  4. Balancing privacy regulations and collaboration: Abiding by data protection regulations, such as GDPR, while promoting a collaborative atmosphere for anti-money laundering initiatives.
  5. Classifying money laundering transactions: Reducing false alarms by accurately differentiating between legitimate and suspicious transactions, thereby minimizing the burden on banks and customers.

Innovative solutions must prioritize data privacy while encouraging collaboration among banks to effectively address these challenges.

Project purpose

This project aims to explore the potential of federated learning in enhancing collaboration between banks for detecting money laundering while preserving data privacy. Additionally, it will examine the prerequisites for effective cooperation between banks and strategize internal processes to tackle associated challenges. The project will investigate employing federated learning to train anti-money laundering models on both synthetic and real transaction data, initially concentrating on synthetic data to showcase the benefits of collaboration.

Taking advantage of AI Sweden's state-of-the-art Edge Learning Lab, the project will be conducted in an advanced, cooperative setting for experimentation and innovation. Simultaneously, the project will consider the regulations and challenges tied to compliance with regard to the General Data Protection Regulation (GDPR). The focus of the project will be on both synthetic and real transaction data, specifically in the context of anti-money laundering efforts. Moreover, it will explore the opportunities for sharing anti-money laundering-related information between banks and external organizations like AI Sweden, pioneering secure information exchange.

Expected outcomes

The project's primary objective is to underscore the importance of collaboration by exhibiting superior model performance in federated learning as opposed to learning in isolation. By fostering cooperation between banks and external organizations, this project will spearhead collaborative anti-money laundering initiatives. The anticipated outcomes include:

  1. Improved detection of money laundering activities across numerous banks.
  2. Advancement of novel approaches within federated learning specifically tailored for anti-money laundering.
  3. Reinforced collaboration between banks and external organizations, tackling shared challenges and nurturing innovation.


Are you interested in learning more about federated learning for anti-money laundering in banking? Reach us here.


The project is funded by Vinnova and coordinated by AI Sweden.
Project partners: Handelsbanken and Swedbank
Project period: February 2023 - February 2025