Digital financial services in Southeast Asia are at an inflexion point, expected to generate revenues of US$38 billion by 2025 and account for 11 per cent of the total financial services industry. Banks and financial services providers are increasingly seeking advanced solutions by leveraging machine learning and AI to tap into this potential.

However, business leaders face two key concerns:

• Making AI operational with reusability, scalability and adaptability with quality data
• Making data privacy business-critical with new techniques of risk management

Solving for dual problems of quality data and data privacy 

A recent survey of 600 data leaders shows that “quality of data” is the top data-related obstacle (42 per cent of respondents) to the adoption of generative AI and large language models. Data privacy and protection (40 per cent) is the second challenge cited by participants. Additionally, researchers also predict that if current Large Language Model (LLM) development trends for training AI models continue, we may run out of available datasets between 2026 and 2032.

Big industry challenges are unlikely to be solved by a single company, working with its proprietary data. When multiple industry players pool their data and collaborate, the collective intelligence generated can solve complex problems such as money laundering, cyber resilience, supply chain management, drug discovery can be tackled more effectively. It’s a win-win situation for individual companies, the industry and customers. Among emerging technologies, Federated Learning (FL) stands out as a revolutionary approach that addresses both concerns: the growing need for data privacy while enabling banks to extract value from distributed data.

Understanding Federated Learning 

Federated Learning is a machine learning paradigm where multiple institutions (e.g., banks) can collaborate to train a shared model while keeping their data decentralised. Unlike traditional machine learning, where data is aggregated into a central location for processing, Federated Learning trains algorithms across decentralised devices or servers holding local data samples, without exchanging them. This approach is particularly appealing to industries like banking, where data privacy and security are paramount. 

Benefits of Federated Learning for operations in banking and financial services 

The benefits of Federated Learning for the banking sector include: 

• Data privacy and security preservation 
  • Protection of sensitive information, as customer information remains within each bank’s secure environment, reducing the risk of data breaches. 
  • Compliance with regulations – such as GDPR, which mandate strict control over personal data and its cross-border transfer.
• Improved model accuracy and robustness
  • Access to diverse data: By leveraging data intelligence from multiple banks, Federated Learning can create more accurate and robust risk management models. This is because the combined data set represents a wider range of scenarios and customer behaviours, leading to better generalisation and prediction capabilities. 
  • Enhanced fraud detection: With access to a broader set of transaction patterns and fraud cases, Federated Learning can improve the detection of fraudulent activities, reducing financial losses.
    
• Efficient resource utilisation 
  • Cost reduction: The Federated Learning approach allows banks to pool their computational resources, reducing the overall cost of model training. This collaborative approach can lead to significant savings in infrastructure and operational expenses. 
  • Accelerated model development: By sharing insights and developments, banks can accelerate the process of model refinement and deployment, leading to quicker implementation of risk management strategies.
    
• Real time risk assessment 
  • Dynamic risk modelling: Federated Learning facilitates the development of models that can be updated in real-time as new data becomes available. This is crucial for identifying emerging risks and adapting to changing market conditions promptly. 
  • Distributed decision making: By enabling localised model updates, more responsive and context-specific decision-making processes within different branches or regions of a bank are supported.
    
• Enhanced collaboration 
  • Cross-institutional collaboration: Banks can collaborate on risk management initiatives without compromising proprietary data, fostering a culture of shared knowledge and best practices within the industry. 
  • Benchmarking and standardisation: Federated Learning enables the creation of industry-wide benchmarks for risk management practices, helping banks to standardise their approaches and improve overall industry resilience.
    
• Regulatory compliance and reporting 
  • Automated reporting: Federated Learning models can be designed to automatically generate compliance reports, ensuring that banks meet regulatory requirements efficiently. 
  • Regulatory sandboxes: Regulators can use Federated Learning to test new policies and regulations on anonymised data sets from multiple banks, assessing their impact without exposing sensitive information. 

Also Read: How Web3 will revolutionise borderless banking in Southeast Asia

 Why Federated Learning is relevant for banking and financial services 

Banks handle vast amounts of sensitive data, including financial transactions, customer information, and behavioural data. This data is not only valuable for making business decisions and improving customer services, but also a prime target for cybercriminals. Moreover, banks operate under strict regulatory frameworks, which impose severe penalties for data breaches or misuse. Federated Learning can enable banks to personalise customer experiences in the following ways: 

• Risk assessment: The Federated Learning collaboration can improve various scoring models by incorporating diverse data from multiple institutions, leading to more accurate assessments of borrowers’ risk profile. When multiple banks shares anonymised and privacy-protected use cases on fraud, threat, risk behaviour, the entire industry benefits from the generated collective intelligence. This sharing of tribal knowledge from each bank, provides insights into industry benchmarks and best practices for local and regional applications to all participants. This further enables banks to understand customer risk profiles and offer relevant products.

• Fraud and money laundering management: Federated Learning intelligence can teach individual bank predictive models, far deeper correlation identifiers for bad actors and bad actions based on private data. This can help identify potential vulnerabilities and mitigating them proactively, so that the customer journey remains free of incident. 

Collective intelligence: The Human Managed architecture for Federated Learning  

In 2018, Human Managed was established in Singapore, to build “collective intelligence” of the crowd – made of humans and machines. Our goal has always been to operate a multi-sided ecosystem-driven platform that gets smarter with more data, more learning and more real world use cases. 

To translate our vision into reality, we created the I.DE.A. (Intelligence Decision Action) platform that builds AI-native solutions for cyber, digital and risk problems for enterprises. This platform is a modular collection of 14 functions and 92 micro-services abstracted into infrastructure, software, data, and AI stacks.  It integrates data from any source, and develops AI models for business context and specific use cases. For individual banks, the platform enables intelligence for smarter decisions and faster actions for better cyber, digital and risk outcomes.  

Integrating data from diverse external sources and generating intelligence in real time, as in the case of risk management for multiple banks, requires privacy preserving technologies. Through Federated Learning and AI-powered apps, the HM collective intelligence platform can build a threat intelligence sharing system for banks that will ensure that:

• Relevant data with the appropriate business context is processed for each bank, via context models 
• Data structures are organised for efficient interoperability or real-time exchange of information across different systems, via STIX v2.1 standards 
• Threat detection for individual banks is agile with cyber incidents detected faster via nano models 
• Privacy across multiple banks is preserved, as a result of collaborative training of machine learning models without actual sharing of raw data via distributed Federated Learning

How it works 

Each participating bank preprocesses its data to ensure consistency and quality before entering the Federated Learning framework. A common initial model is shared among the banks, which will be locally trained on their respective datasets. Each bank trains the model locally on its own data, generating model updates (e.g., gradients).

Also Read: Gen AI in banking: How to ensure a successful transformation for an age-old industry

The local updates are securely aggregated using techniques like secure aggregation protocols or homomorphic encryption. The aggregated updates are used to refine the global model, which is then redistributed to the banks for further training. These steps are repeated iteratively until the model converges to an optimal state. 

Conclusion: The future of intelligence is collective 

The future of effective, real-time intelligence will need to be based on collaborative efforts. Federated Learning can be leveraged in banking to enhance services, improve decision-making, and ensure compliance with stringent data protection regulations.

Overtime, we believe that Federated Learning will drive digital transformation in banking and level the playing field for banks of all sizes. It will foster innovation and create new business models. It will allow for greater financial inclusion, with a greater number of people, especially the rural unbanked access services for personal and business needs.

—

Editor’s note: e27 aims to foster thought leadership by publishing views from the community. Share your opinion by submitting an article, video, podcast, or infographic.

Join us on Instagram, Facebook, X, and LinkedIn to stay connected.

Header image credit: Canva Pro

The post The role of Federated Learning in enhancing financial services in Southeast Asia appeared first on e27.